Dual N-Back: What the Research Actually Shows

Few tasks in cognitive psychology have generated as much excitement — and as much subsequent skepticism — as dual N-Back. The 2008 study by Susanne Jaeggi and colleagues at the University of Michigan suggested that a few weeks of N-Back practice could raise fluid intelligence, a claim that, if true, would overturn decades of consensus. The study triggered an explosion of N-Back-based apps and a wave of replication attempts. More than fifteen years later, where does the evidence stand? This article tries to give an honest answer.

What N-Back is

In a standard N-Back task, you see a stream of stimuli — usually letters appearing in different positions on a 3×3 grid — and you must press a button whenever the current item matches the one that appeared N steps back. At N=2, you respond if the current item matches the one from two trials ago. At N=3, you track three trials back. The dual version adds a second stream: usually the letter's identity is one stream, and its position is the other, and you respond to matches in each independently.

The task is demanding because the contents of working memory must be continuously updated. Every trial, the oldest item drops out, every other item shifts back one position, and a new item enters. You cannot just hold a static list; you have to keep refreshing it. This is closer to what working memory does in real life than the static storage measured by digit span.

The 2008 Jaeggi study

Jaeggi et al. recruited 70 healthy young adults and assigned them to either practice dual N-Back for up to 19 days (with the difficulty level — N — increasing as they improved) or to a no-contact control group. Both groups took a test of fluid intelligence (Raven's Progressive Matrices) before and after the training period. The training group showed a statistically significant gain in fluid intelligence scores, and the gain was larger for participants who trained for more days. The authors concluded that working-memory training could produce transfer to fluid intelligence.

Published in PNAS, the study was widely covered in the press. It suggested something close to a cognitive free lunch: a few weeks of a computer game producing gains on an IQ-like measure. Within a couple of years, a small industry of N-Back apps had sprung up to commercialize the finding.

The replication problem

Science works by replication. A single study, however well-designed, can be a fluke — a positive result that arises from sample-specific noise, an unintended confound, or publication bias. The right response to an exciting finding is to try to replicate it independently, with larger samples and tighter controls. In the case of N-Back, this has been done many times.

The results have been mostly disappointing. A 2013 meta-analysis by Melby-Lervåg and Hulme examined 23 working-memory training studies and concluded that while training produced reliable short-term gains on the trained task, there was "no convincing evidence" of transfer to other cognitive measures, even closely related ones. A 2018 cross-sectional study by Sala and Gobet, published in PMC, reached a similar conclusion: brain-training interventions, including N-Back, produced essentially no transfer to untrained cognitive tasks. Some smaller studies have found positive transfer, but the overall weight of evidence has shifted toward skepticism.

This is not the same as saying the 2008 study was wrong. It might be partially right — there may be a small transfer effect that subsequent studies with less statistical power failed to detect. But the strong version of the claim ("N-Back raises fluid intelligence") is not something a careful reader should accept on the basis of the available evidence.

What N-Back practice definitely does

• Improves performance on N-Back tasks. This is uncontested. A few weeks of practice will raise the N level you can handle.
• Improves performance on closely related working-memory tasks. There is some evidence of "near transfer" to tasks that share cognitive components with N-Back.
• Does not reliably improve fluid intelligence. This is the contested claim that has mostly failed to replicate.
• Does not reliably improve broad cognitive ability. The "far transfer" claim is the one the literature most strongly rejects.

Why the original finding might have been overstated

Several methodological issues have been raised about the 2008 study and the literature that followed it. First, the control group was a no-contact control — participants who simply waited, then retook the test. A better design uses an active control group that does some unrelated cognitive task (like crossword puzzles) for the same amount of time, to control for the general effect of "spending time thinking." When active controls are used, the apparent transfer effect tends to shrink or disappear.

Second, IQ tests are subject to practice effects: taking the test twice produces a small gain even without any intervention. If the training group practices a demanding cognitive task daily and the control group does nothing, the training group may simply be more "warmed up" cognitively at retest — an effect that has nothing to do with N-Back specifically.

Third, publication bias may have inflated the apparent effect size. Studies that find positive results are more likely to be published than studies that find null results, so the published literature over-represents positive findings. Meta-analyses that correct for publication bias tend to find smaller effects than meta-analyses that don't.

Should you practice N-Back?

If you find the task engaging, there is no reason not to. It exercises working memory in a way that few other tasks do, and the practice effect on the task itself is real. What you should not expect is a measurable boost to your general cognitive ability. If a website or app tells you N-Back will make you smarter, treat that claim with skepticism. The current evidence does not support it.

You can try the task yourself with our Dual N-Back tool. Be prepared: it is genuinely hard, especially at N=3 and above. Most people find N=2 a good starting point and progress to N=3 over a few sessions.

Sources

• Jaeggi, S. M., et al. (2008). Improving fluid intelligence with training on working memory. PNAS, 105(19), 6829-6833.
• Melby-Lervåg, M., & Hulme, C. (2013). Is working memory training effective? A meta-analytic review. Developmental Psychology, 49(2), 270-291.
• Sala, G., & Gobet, F. (2018). Does far transfer exist? A cross-sectional investigation into the efficacy of brain training. PMC. pmc.ncbi.nlm.nih.gov/articles/PMC6629869
• PMC (2021). Dual n-back working memory training evinces superior transfer effects. PMC link