Dual n-back is a memory exercise in which you track two streams at once — a square jumping around a grid and a voice reading letters — and signal whenever either one matches what came n steps earlier. It became famous after a 2008 study reported that practising it raised fluid intelligence. People do get much better at dual n-back with practice; the larger trials and meta-analyses since then find that the improvement stays largely with the task and does little for IQ.
What the n-back task is
In the basic n-back task you watch or hear a sequence of items, one every few seconds, and respond whenever the current item is the same as the one shown n steps before. The task was first described by Wayne Kirchner in 1958, in a study of age differences in holding on to rapidly changing information. With letters, it looks like this:
| Level | Sequence | Respond on |
|---|---|---|
| 1-back | T · K · K · R · B | the second K (same as one step ago) |
| 2-back | T · K · T · R · B | the second T (same as two steps ago) |
| 3-back | T · K · R · T · B | the second T (same as three steps ago) |
Each new item has to be compared with an old one, then stored, while the oldest one is dropped. That constant updating is what makes the task a load on working memory, and why difficulty climbs steeply with every step of n.
What makes it “dual”
Dual n-back runs two n-back tasks at the same time. A square appears in one of the positions on a grid while a voice speaks a letter, and you make two separate judgements on every step: does the position match the one from n steps ago, and does the letter? This is the version used in the 2008 training study, and most apps copy it. Typically the level rises when you are accurate and drops when you are not, so the task stays hard however much you practise.
How to do dual n-back
- Start at 1-back. Compare each position and each letter with the one immediately before it.
- Use two responses. One key for a position match, another for a letter match, both when both match, nothing when neither does.
- Move up when you are accurate. At 2-back you compare with two steps ago, at 3-back with three. Holding both streams at 3-back is already demanding for most people.
- Expect to plateau. Scores rise quickly at first and then level off; that curve is normal for any new skill.
How researchers use the n-back
Long before it was sold as training, the n-back was a laboratory tool. It is one of the most common tasks in brain imaging studies of working memory: a 2005 meta-analysis of 24 imaging studies found that n-back variants consistently engage a network across the frontal and parietal lobes, including the dorsolateral prefrontal cortex.
As a measure of an individual person, though, it is weaker than it looks. A 2007 study found that n-back performance correlated only weakly with a standard working memory span task, suggesting the two do not mainly measure the same thing — yet each explained its own share of fluid intelligence. A 2010 study by the team behind the 2008 training paper concluded that the n-back is not a useful measure of individual differences in working memory, partly because its reliability is too low, while noting that at higher levels of load it does predict fluid intelligence.
The 2008 study that started the craze
Susanne Jaeggi and colleagues at the University of Bern and the University of Michigan trained adults on dual n-back and tested them on a fluid reasoning test before and after. They reported transfer to the reasoning test, even though the test looked nothing like the training task, and that the gain depended on the amount of training: the more sessions, the larger the improvement. Published in PNAS, it was the first strong claim that fluid intelligence could be raised by practising something other than the test itself, and it set off a wave of brain-training apps and headlines.
What happened when others tested it
- 2013, a randomised trial with a placebo. Thomas Redick and colleagues gave young adults 20 sessions of adaptive dual n-back, compared them with an active placebo group (adaptive visual search) and a no-contact group, and tested fluid intelligence, working memory, multitasking, crystallised knowledge and perceptual speed. Both groups improved on their own training task. Despite high statistical power, there was no transfer to any of the ability tests. The authors had argued that earlier positive studies often compared training with a group that did nothing, which cannot separate real transfer from expectation and effort.
- 2015, a meta-analysis of 20 studies. Jacky Au and colleagues, including two authors of the 2008 paper, pooled n-back studies in healthy adults and reported a small but significant positive effect on fluid intelligence.
- 2016, middle-aged adults. A randomised trial of 57 adults aged 30 to 60 compared five weeks of online dual n-back with processing-speed training and found no improvement in working memory, processing speed or fluid intelligence; Bayesian analyses supported the null result.
- 2016, the widest view of working memory training. A meta-analysis of 87 publications and 145 comparisons found reliable short-term gains on working memory tasks but no convincing evidence of far transfer when training was compared with an active control group — and its publication-bias analysis found no evidential value in the studies that used active controls.
- 2017, n-back studies only. A multi-level meta-analysis of 203 effect sizes from 33 randomised controlled trials separated the kinds of gain. Transfer to untrained n-back tasks was medium-sized; transfer to other working memory tasks, fluid intelligence and cognitive control was very small. Neither the amount of training nor single-versus-dual versions changed the result.
A 2016 review of brain training in general, written for policymakers, reached the same pattern: extensive evidence of improvement on the trained task, less on closely related tasks, and little evidence of benefits for distantly related tasks or everyday life.
So can dual n-back raise your IQ?
The most defensible reading is that dual n-back makes you better at dual n-back, somewhat better at similar tasks, and at most marginally better on reasoning tests — an effect that shrinks further when the comparison group also does something engaging. The two pooled analyses that found a small effect on fluid intelligence and the ones that found almost none are measuring the same small number from different angles; none supports the idea of lasting, meaningful IQ gains from a few weeks of practice.
That fits what the brain actually does: it changes with practice, but the change is specific to what is practised — see neuroplasticity. The wider question of what does and does not move IQ scores is on our page on whether you can increase your IQ.
Dual n-back and ADHD
Because working memory difficulties are common in ADHD, n-back and similar programs are often marketed to people with ADHD. A 2015 meta-analysis of 16 randomised trials with 759 children found that cognitive training improved scores on laboratory working memory tests, but working memory training specifically had no effect on ADHD symptoms, and the benefits reported by parents shrank when the raters were likely blind to the treatment. This page does not offer treatment advice; decisions about ADHD belong with a qualified professional.
Is it worth doing?
As a demanding game, dual n-back is harmless and some people enjoy the challenge. As a route to a higher IQ, the evidence does not support it. If the goal is a specific skill, practising that skill beats practising an abstract task; if the goal is managing everyday memory load, writing things down and breaking tasks into steps help more reliably, as explained on our page on executive function.
Trust and scope notes
This page summarises published research in general terms. It is not a diagnostic tool, does not assess ADHD or any other condition, and does not recommend or review any training product. Our IQ test measures reasoning; it does not include an n-back task or report a working memory score.
IQ Revealed is not affiliated with, endorsed by or connected to the researchers, universities, journals or app makers named or referred to on this page. Their names are used only to identify the work being described.