Dual N-Back

How to play dual n-back

Dual n-back asks you to follow two things at once, where a square appears and which letter you hear, and to notice when either one repeats from a set number of trials back. That number is N. This is a whole session, from start to finish.

1. Press Play on the card above. The game opens over this page, and Close brings you back here with your session paused.
2. The first time you play, the game offers four short warm-ups before your first real session. Take them, or skip straight to 2-back, where you compare each trial with the one two trials earlier.
3. Press START. A new trial comes every 3 seconds: one square of the grid lights for half a second, and a voice says one letter.
4. Press A or Position match when the square is in the same place as N trials ago.
5. Press L or Sound match when the letter is the same as N trials ago.
6. When both match, press both. When neither matches, press nothing. Not pressing is how you answer "no match".
7. The first N trials have nothing before them to compare with. The buttons rest, so just remember what you see and hear.
8. After 20 + N trials, 22 at 2-back, you get your score, the level of your next session and a review of every trial.

The four warm-ups

The first time you play, the game splits the task into four steps of 10 trials each, about 30 seconds apiece, so you meet one idea at a time:

• Step 1, 1-back, squares only: press A when the square repeats the one just before.
• Step 2, 1-back, letters only: press L when the letter repeats the one just before.
• Step 3, 1-back, both together: squares and letters, still against the one just before.
• Step 4, 2-back, both: the real task, compared with two trials back.

In steps 1 to 3, training wheels tint the square you are comparing with and show its letter above the grid, and a coach line under the grid explains each trial as it ends. Step 4 starts with the wheels off, and the real session has none. Any step can be skipped, the warm-ups never count toward your level or your progress, and Replay the warm-up under START brings the steps back later.

Controls

Feedback while you play

A press turns its button green at once if it was a match, and coral if it was not. A miss can only be known when the trial ends, so the line above the buttons sums up each trial in words, such as "Trial 7: position hit, sound missed." Only the lit square and the buttons change color. Nothing flashes.

If you like to see the whole session at a glance, turn on the trial strip in Settings. A row of small tiles under the grid then fills in one column per trial: a check for a hit, a ring for a miss, a cross for a false alarm and a dot for a correct rest. It is off unless you turn it on. Every session ends with a review either way, where you can select any trial to see the one it was compared with.

With or without sound

Sound is one setting shared by every game in the arcade, and it starts off. If you have never set it, pressing START turns it on, since the letters are spoken. If you turned it off, it stays off. Each letter then appears as text in the middle square of the grid, the one square that never lights, and a note at the top of the game says so, with a Turn sound on button.

What N means

N is how far back you compare. At 1-back each trial is checked against the one just before it. At 2-back it is checked against the one before that, so you carry the last two squares and the last two letters and update them every 3 seconds.

A worked example at 2-back

Trial	Square	Letter	Compared with	Position	Sound
1	top left	K	Nothing yet	–	–
2	middle right	R	Nothing yet	–	–
3	top left	T	Trial 1	Match	No
4	bottom middle	R	Trial 2	No	Match
5	top left	T	Trial 3	Match	Match
6	top left	C	Trial 4	No	No

Trials 1 and 2 have nothing two back, so they only have to be remembered. Trial 3 is on the same square as trial 1, a position match. Trial 4 says the same letter as trial 2, a sound match. Trial 5 repeats trial 3 on both. Only the trial exactly N back counts: trial 6 sits on the same square as trial 5, which would be a match at 1-back but is not one at 2-back.

Scoring and levels

Rules is a setting with two choices. Both come from Brain Workshop. They count a session differently and move the level at different marks. Classic is the default.

Classic rules, Brain Workshop's default

• The score is hits ÷ (hits + misses + false alarms), with both streams added together. A trial with no match and no press counts neither way.
• 80% or more moves N up by one.
• 50% to 79% keeps the level.
• Under 50% is a strike. Three strikes at one level, in a row or not, take N down by one. Any level change clears the strikes.

Research rules, Brain Workshop's Jaeggi mode

• Each stream is scored on its own, as (hits + correct rests) ÷ scored trials. A correct rest, no press when nothing matched, counts as right. The session's score is the lower of the two streams.
• 90% or more moves N up by one.
• 75% to 89% keeps the level.
• Under 75% moves N down by one after that one session.

Brain Workshop named this mode after the 2008 study, but the study itself moved the level by counting mistakes: up after fewer than three in each stream, and down after more than five. Echo Invaders on this site moves its level at the same 90% and 75%.

The words in a score

• Hit: you pressed, and it was a match.
• Miss: it was a match, and you did not press.
• False alarm: you pressed, and it was not a match.
• Correct rest: it was not a match, and you pressed nothing.

Under Classic rules a correct rest is left out of the score. That keeps the score honest. Pressing nothing all session scores 0%, and pressing on every trial scores poorly too, since every wrong press counts against you. Two examples:

• You catch all 12 matches and press 3 times when nothing matched: 12 ÷ (12 + 0 + 3) = 80%. That is just enough to move up.
• You catch 9 matches, miss 3 and make no false alarms: 9 ÷ (9 + 3 + 0) = 75%. You stay at your level.

One session, two verdicts

Here is one 2-back dual session scored both ways. The player caught every match but pressed Sound match three times when the letter had not repeated.

• Classic rules: 12 hits ÷ (12 hits + 0 misses + 3 false alarms) = 80%, up to 3-back.
• Research rules: position 20 ÷ 20 = 100%, sound 17 ÷ 20 = 85%. The lower score, 85%, is the session's, so it stays at 2-back.

The result screen shows the score the way your rules count it, with a bar for each stream of its hits, misses and false alarms, so you can see whether the squares or the letters cost you more. It also says what the other rule set would have made of the same session. N never goes below 1, and with a fixed level the score is shown but N stays where you set it.

What a session is

A session is 20 + N trials: 22 at 2-back, which takes about 66 seconds, and 23 at 3-back. Of the 20 trials you answer, 6 are position matches and 6 are sound matches, and 2 of those fall on the same trial. This is the design of the blocks in the 2008 study. The game also makes sure that no other trial matches by chance. Brain Workshop's own default session is longer, 20 + N squared trials (24 at 2-back and 29 at 3-back), and Session length in Settings offers it too. Brain Workshop also deals its matches at random, so their number changes from session to session there.

High scores

Each session scores 100 points for each N plus its percent, so 85% at 3-back is 385, and the high scores rank the highest N first. A session under 50% is never a best: it counts in your progress and your saved results, but not in the high scores. Your best sessions are kept in this browser. Logged in to the forum, you can show them in the high scores on the forum and the site. Only sessions played on the default game settings, with the letters heard rather than read, are shown there, so every row means the same task.

Settings and modes

Settings is on the game's start screen and under the grid. Its defaults are the classic task, so you can leave them alone and press START. Opening it during a session pauses the session. Game settings apply from your next session, and sound and display settings change at once. Your choices are kept for next time.

What you follow

• Dual n-back: squares and letters together, two answers. The classic task, and the default.
• Position n-back: single n-back with the grid alone and one wide answer button. There are no letters, so it needs no sound.
• Sound n-back: single n-back with the spoken letters alone and one wide answer button. The squares stay dark.

A single mode is a gentler way into the task if dual feels like too much at first. Playing each one at the same level also shows you which of the two streams is harder for you.

Every setting

Sound and display

Sound check plays one letter so you can be sure the voice comes through before you start. If you did not hear it, the game offers to show the letters as text instead. Captions are for players who are hard of hearing. With captions on and sound on, both streams are on screen, a slightly different task from the classic one, where you see one stream and hear the other, so these sessions are marked in your results. The letters are recordings, the same in every browser, and the game never says "capital C".

Dual n-back training: how much and how often

There is no agreed amount. What there is, is the schedule the best-known study used, and it makes a useful reference point.

The 2008 schedule

In the study by Jaeggi and colleagues, young adults trained on dual n-back for about 25 minutes a day, with each day's training made of 20 blocks, and the groups trained for 8, 12, 17 or 19 days. A session here is one of those blocks, the same length as theirs. At 2-back, 20 sessions are 22 minutes of trials, plus the short pauses between them, so one study day fits in about the same time here.

A routine you can keep

These are suggestions for playing well, not findings from the studies.

• Choose a number of sessions you can manage on most days, such as 10, or one full study day of 20.
• Play somewhere you can hear the letters, or turn sound off and read them in the middle square.
• Take a break when your scores start to slide. A tired session tells you little about your level.
• Judge your progress by the week, not by the session. One session can go well or badly for many reasons.

Getting better at the task itself is the one result the studies agree on, as the research section below explains. Progress, beside Train in the game, keeps your log: each day's average and best N over the last three weeks, a calendar of the last eight weeks and your totals. Turn on the daily set of 20 on its Today card, or in Settings, to count your sessions toward one study day.

The n-back test

The n-back task began as a test, not a game. Wayne Kirchner introduced it in 1958, in a study of age differences in short-term memory for information that changes quickly. Used as a test, it measures working memory, the small amount you can hold and update in your head from moment to moment: the same task, played once at set levels, rather than every day to get better at it.

How to test yourself here

Take the level test from the game's start screen. It plays three dual rounds in a row at 1-back, 2-back and 3-back, 21, 22 and 23 trials, with no feedback until the end, so the rounds stay comparable from one sitting to the next. Your level today is the highest round you scored 80% or more in, and training then starts from it. Under 1-back, training starts at 1-back.

Treat the number as a snapshot of one day. With 12 matches in a round, a single slip can move the result by a level. This page has no norms to compare it with, so compare it only with your own earlier results, which Progress lists, and take the test again every few weeks to see your trend. It is not an IQ test, and a level here is not an IQ score.

What the research says

• 1958: Kirchner introduces the n-back task, in a study of age differences in short-term memory for information that changes quickly.
• 2008: Jaeggi and colleagues make dual n-back famous. Young adults trained about 25 minutes a day for 8, 12, 17 or 19 days. Their fluid intelligence scores, on matrix reasoning tests, rose more than those of groups who did no training, and more with more days of training. The untrained groups improved too, but less.
• 2013: Redick and colleagues test it against an active control. The comparison group practiced an adaptive visual search task instead. Both groups got better at what they practiced, and neither showed any gain in fluid intelligence, multitasking, working memory capacity, crystallized intelligence or perceptual speed.
• 2013: Melby-Lervåg and Hulme pool 23 studies of working memory training. They found short-term gains on working memory tasks like the ones trained, but no convincing gains in other skills such as verbal and nonverbal ability, word decoding or arithmetic. Verbal working memory gains had faded at follow-up.
• 2015: Au and colleagues pool 20 n-back studies. They found a small positive effect on fluid intelligence, about 3 to 4 IQ points by the authors' estimate. It was larger in studies whose control groups did nothing than in studies with active control groups, though the authors argue that the kind of control group does not explain the difference.
• 2016: Melby-Lervåg, Redick and Hulme review 87 publications. Working memory measures improved reliably right after training, there was no convincing transfer to other abilities when trained groups were compared with control groups given another task, and a bigger working memory gain did not go with more transfer.
• 2017: Soveri and colleagues look at 33 randomized n-back training studies. They found medium-sized transfer to untrained n-back tasks, and very small effects on other working memory tasks, cognitive control and fluid intelligence.

In short, dual n-back training reliably makes you better at dual n-back and at tasks close to it. Whether it raises intelligence is disputed, and any effect found has been small. Play it as a demanding focus and working memory exercise, not as a way to raise your IQ.

Sources

• Kirchner, W. K. (1958). Age differences in short-term retention of rapidly changing information. Journal of Experimental Psychology, 55(4), 352–358.
• Jaeggi, S. M., Buschkuehl, M., Jonides, J., & Perrig, W. J. (2008). Improving fluid intelligence with training on working memory. PNAS, 105(19), 6829–6833.
• Redick, T. S., et al. (2013). No evidence of intelligence improvement after working memory training: A randomized, placebo-controlled study. Journal of Experimental Psychology: General, 142(2), 359–379.
• Melby-Lervåg, M., & Hulme, C. (2013). Is working memory training effective? A meta-analytic review. Developmental Psychology, 49(2), 270–291.
• 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.
• 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": Evidence from a meta-analytic review. Perspectives on Psychological Science, 11(4), 512–534.
• Soveri, A., Antfolk, J., Karlsson, L., Salo, B., & Laine, M. (2017). Working memory training revisited: A multi-level meta-analysis of n-back training studies. Psychonomic Bulletin & Review, 24(4), 1077–1096.

Dual n-back FAQ

What is dual n-back?

Dual n-back is a working memory task with two streams at once. On each trial a square lights on a 3 by 3 grid and a voice says a letter, and you respond whenever the square or the letter matches the one from N trials earlier.

Do I have to do the warm-up?

No. Skip to 2-back from the first screen, or skip any single step. The warm-up is there because 2-back with squares and letters at once is hard to follow the first time. Replay the warm-up under START brings it back whenever you want, and the warm-ups never count toward your level.

Does dual n-back increase IQ?

Probably not, or only a little. A 2008 study reported gains in fluid intelligence, but later controlled studies and meta-analyses found that players mostly get better at n-back and similar tasks, with any effect on intelligence small.

How long should you do dual n-back each day?

There is no agreed amount. The best-known study used about 25 minutes a day, 20 blocks a day for 8 to 19 days, which is about 20 sessions at 2-back in this game. A shorter routine you keep up is a fair start.

What is a good dual n-back level?

A good level is one you can hold at 80% or more under the Classic rules, since that is where the game moves you up. There are no norms here to compare with, so track your own level over weeks on the Progress tab.

What is the difference between n-back and dual n-back?

Single n-back has one stream to follow, such as letters only. Dual n-back has two at once, a square and a letter, and you answer each one separately. Position n-back and Sound n-back in Settings are the two single forms, and Echo Invaders on this site is a single n-back game with letters.

Can you play dual n-back without sound?

Yes. With the sound off, each letter appears as text in the middle square, the one square that never lights, and a note at the top of the game says so. Sound is one setting for every arcade game on the site and it starts off. The first time you press START it comes on, since this game speaks its letters, but if you turned it off yourself it stays off until you turn it back on.

Is there a dual n-back for deaf players?

This one can be played without hearing anything. Turn on Captions, or leave the sound off, and every letter is shown in the middle square. Both streams are then on screen, which is a slightly different task from the classic, so sessions played this way are marked in your results.

Does dual n-back work with a screen reader?

Yes. Each trial is announced as a short message such as "Trial 5 of 22: top right, K", so the square becomes a spoken position. The answer buttons give their keys and report themselves unavailable during the first N trials, when there is nothing to compare yet.

Can I play dual n-back with a keyboard only?

Yes. Press A for a position match, L for a sound match, and Space or Escape to pause. Tab reaches every button and switch, and the focus ring stays visible. Some screen readers use single letters to move around a page, so with one running you may prefer to Tab to the answer buttons and press Enter.