Dual n-back works for what it trains: you get much better at n-back and somewhat better at similar working memory tasks. Claims that it raises IQ are weak. The best-known meta-analysis found about 3 to 4 IQ points on average, and studies with an active control group show close to zero.
What is dual n-back?
Dual n-back is a working memory task with two streams at once. At each step, a square lights up in a grid and you see or hear a letter, and you respond when the position or the letter matches the one from N steps back.
At 2-back, you compare every new square and letter with the ones from two steps earlier, then update what you are holding. That constant compare-and-update cycle is what makes the task hard. You can try it free on our dual n-back test, with 1-, 2- and 3-back levels and rounds of about a minute.
Where did the n-back task come from?
Wayne Kirchner introduced the n-back task in 1958 to study short-term memory. His paper in the Journal of Experimental Psychology looked at age differences in remembering "rapidly changing information", and it used a single stream of items.
For the next five decades researchers used n-back mainly as a measuring tool, to load working memory in experiments. A 2005 meta-analysis by Adrian Owen and colleagues pooled 24 brain-imaging studies that used it.
What did the 2008 Jaeggi study find?
The 2008 study by Susanne Jaeggi, Martin Buschkuehl, John Jonides and Walter Perrig, published in PNAS, reported that dual n-back training raised scores on a fluid intelligence test. Fluid intelligence is the ability to solve new problems, usually measured with matrix puzzles.
The design, from the paper:
- Participants: healthy young adults, 34 trained and 35 controls
- Training: about 25 minutes a day, on weekdays
- Groups: 8, 12, 17 or 19 days of training
- Pace: one square and one spoken letter every 3 seconds
- Test: Raven's matrices or BOMAT, with a 10-minute time limit
- Result: trained groups gained more than controls, and longer training gave bigger gains
The dose-response pattern drew most of the attention. "The more training, the more improvement in fluid intelligence," Jaeggi said in the University of Michigan press release.
Critics found problems quickly. In 2009 Moody pointed out that the BOMAT matrix test ran with a 10-minute limit instead of the usual 45 minutes, so the scores may reflect speed more than reasoning. The controls received no training of any kind, so expectations and motivation could explain part of the gap.
Did later studies replicate it?
Later studies replicated the gains on n-back itself, while the gains in intelligence came and went. A 2010 study by Jaeggi's group found that single n-back, with only visual squares, produced gains in fluid intelligence (Gf) almost equal to dual n-back.
Independent teams were less positive. Redick and colleagues (2012) and Chooi and Thompson (2012) both trained people on n-back and found no gain in fluid intelligence.
What do the meta-analyses say?
The meta-analyses agree on one point: well-controlled studies show small or no effects on intelligence. They differ on whether the small effect is real.
- Melby-Lervåg and Hulme, 2013: covered 23 working memory training studies. Main finding: short-lived gains on similar tasks, no lasting far transfer.
- Au et al., 2015: covered 20 n-back studies in adults. Main finding: a small effect on fluid intelligence, g = 0.24, about 3 to 4 IQ points.
- Soveri et al., 2017: covered n-back training studies. Main finding: the biggest gains on untrained versions of the n-back task, and small effects on other working memory tasks, cognitive control and Gf.
Melby-Lervåg and Hulme (2013) looked at working memory training in general, not only n-back. Gains on close tasks faded over time, and a modest effect on non-verbal ability disappeared once they kept only the better-designed studies.
Au and colleagues (2015), a team that included Jaeggi, reported the 3 to 4 IQ point estimate. Their own data show why it is contested: studies with passive control groups, where controls did nothing, averaged g = 0.44, while studies with active control groups, where controls did another task, averaged g = 0.06. A Bayesian re-analysis by Dougherty and colleagues (2015) found that the active-control studies favour no effect, and concluded that the evidence largely fails to support IQ gains from short-term n-back training.
Soveri and colleagues (2017) used a stricter definition of near transfer, counting only untrained versions of the n-back task. Those versions improved the most. Transfer beyond them was small, and the authors suggest people learn strategies specific to the task more than they expand memory capacity.
An independent meta-analysis by the writer Gwern Branwen reached the same split: an overall d of about 0.35, but about 0.14 in active-control studies, with a confidence interval that includes zero.
Why does the control group matter so much?
The control group matters because people who know they are in a training study often try harder at the post-test. A group that does nothing has no reason to expect improvement, so the trained group looks better even if the training changed nothing.
When controls play a different game for the same time, that expectation effect is shared. Across the n-back studies, the IQ effect shrinks to near zero under those conditions. That pattern is the main reason most researchers now doubt large IQ claims.
What are the real benefits of dual n-back?
The solid benefit is a skill: holding a few items in mind, comparing them with new input and updating them at a steady pace. Your n-back level will rise with practice, and some of that carries to similar tasks.
That skill has value on its own terms. You use the same compare-then-update routine when you follow a fast conversation or keep a running total. Treat any change in reasoning or IQ as a possible bonus. Our science page labels N-Back as "B, core process" for that reason: gains inside the task are robust, carry-over is debated.
How should you practise dual n-back?
Start at 2-back, move up once you pass about 80% accuracy and step down if you fall below about 50%. Those thresholds come from our free dual n-back test, where each round has 20 scored steps and takes about a minute.
A few habits make practice count:
- Keep a short queue. Hold only the last N items as a chant, such as "left, top, centre". When a new item comes, drop the oldest and add the newest.
- Compare, then update. First ask "same as N ago?", then push the new item into the queue. Doing both at once causes most errors.
- Watch false alarms. The score counts presses on non-matches against you, so pressing at every step scores low. Only press when you are sure.
- Compare scores at the same level. A new level feels hard for the first few rounds. Judge progress at a fixed N.
For session length, the studies give a rough guide rather than a rule. Jaeggi's participants trained about 25 minutes a day, and the shortest sessions in the Au meta-analysis were 18.5 minutes. In Au's data, the number of sessions did not predict transfer. If your goal is the skill, 10 to 20 minutes of focused rounds, most days, is a reasonable plan. Stop when tiredness starts to pull your accuracy down.
In the Omni Brain app the same idea runs as the N-Back game, where difficulty adapts after every block to keep you near 85% correct, so you don't need to manage levels yourself.
Questions about dual n-back
Does dual n-back increase IQ? The evidence is weak. One meta-analysis estimated 3 to 4 IQ points, but studies where the control group did another activity show close to no effect. Expect better n-back scores, and treat IQ gains as unproven.
What are the benefits of dual n-back? You get better at holding and updating information under time pressure, and some of that transfers to similar working memory tasks. Broader benefits for reasoning, school or work have not held up in well-controlled studies.
How long should I play dual n-back each day? Most training studies used 20 to 25 minutes a day over several weeks. No study has found an ideal length, and more sessions did not predict more transfer in the Au meta-analysis. Ten to twenty focused minutes most days is enough to improve at the task.
Is single n-back as good as dual n-back? In a 2010 study by Jaeggi's group, single n-back with visual squares produced gains in fluid intelligence almost equal to dual n-back. If two streams feel overwhelming, single n-back is a fair place to build the habit.
What is a good dual n-back score? On our dual n-back test, above 80% at a level means you are ready to move up, and below 50% means step down. Compare yourself at the same N, because each level up is a large jump in difficulty.
Practise n-back every day
Omni Brain's N-Back game adapts to you as you play, and 81 more games train memory, attention, speed and other skills, each with a technique you can use outside the app. The free plan includes a daily workout of three games in about five minutes. Download Omni Brain, or start right now with the free dual n-back test in your browser.