Does Dual n-Back Make You Smarter? What the Evidence Actually Shows

Dual n-back sounds like a party trick for your memory. Two streams of information, a running count in your head, and a button to press when something matches. People have trained on it for years hoping to get smarter. Whether that works is a more interesting story than the game itself.

What the Dual n-Back Task Asks You to Do

Dual n-back is a working memory task. According to the Gwern.net meta-analysis page, it was developed for cognitive testing batteries and only later repurposed for training. Its core demand is simple to state. Hold several items in mind at once, and keep updating them as new ones arrive.

The hard part is the letter n. Every new item has to be compared with the one shown n steps earlier. You do that while also watching a second stream of information. Raise n and the memory load climbs, because you now carry more items at the same time.

One caveat. The sources I have do not describe the screen layout, the stimulus types, or how levels progress in a session. If you want that practical detail, you will need a separate guide to how the task is run.

The Original Claim and the Study That Started It

The claim that made dual n-back famous came from Jaeggi and colleagues in 2008. As summarized on Gwern.net, that first training study reported that practicing dual n-back raised IQ test scores in healthy young adults. That is why people talk about it as a way to boost general intelligence, not just memory.

Other groups tested the idea. One example is a PLOS ONE paper by Linette Lawlor-Savage and Vina Goghari, titled Dual N-Back Working Memory Training in Healthy Adults: A Randomized Comparison to Processing Speed Training. I can confirm the title. I could not read the text in the version available to me, so I will not guess at its methods, sample size, or results.

What we can say is modest. The idea has a clear origin, and more than one team has put it to the test. Whether the gains hold up is a separate question, and the answer turns on how each study was built.

What a Meta-Analysis Found About Study Design

The most useful source here is a meta-analysis on Gwern.net. It covers more than 19 studies, published up to 2016, that measured IQ after an n-back intervention. Across all of them, the post-training IQ gain is described as medium-sized. Note that this is one author's analysis on a personal site. The source does not say whether it was peer-reviewed or published in a journal.

The interesting finding is what predicts the size of the gain. It tracks closely with the type of control group. Studies with passive controls, where the comparison group did nothing, show bigger gains. Studies with active controls, where the comparison group trained on something else, show only a small positive effect.

The author argues that big passive-control gains may come from control participants trying less hard on the post-test, not from a real rise in fluid intelligence. For the small leftover effect, the author floats three explanations: training specific to matrix-style tests, undetected publication bias, and smaller motivational effects. The bottom line on the page is blunt. The claim that dual n-back raises fluid intelligence is rated probably not true, with the certainty called highly likely.

How to Judge a Claim About Brain Training

You do not need a statistics degree to read these claims sensibly. Start with the control group. If the comparison group trained on another task, you can separate the training effect from the effect of just being in a study. If they did nothing, motivation and effort remain live explanations.

Next, ask whether the testing was set up so that motivation differences could not explain the result. Then ask what is actually being claimed. Getting better at dual n-back is one thing. Getting better at reasoning in general is another.

Watch for a few red flags:

• Big IQ gains that come from passive-control designs.
• Claims built on a few studies with large gains while the wider literature, which also includes studies with no significant gain, gets ignored.
• Promises of lasting, real-world benefit. The sources do not address whether any transfer lasts past the post-test or reaches everyday tasks.

What the Sources Do Not Cover

This piece has limits, and they are worth stating plainly. The sources give no guidance on choosing a program or app. They describe how study design shapes results, which is not the same as a buying guide. They also give no cost or time-commitment figures, so anyone planning a schedule needs to look elsewhere.

Two of the three sources also failed me. The PLOS ONE text was unreadable in the version I had, and the Springer link returned only a browser-challenge page. So the evidence here rests mainly on one meta-analysis and its summary of the 2008 study. Treat the conclusions as a careful read of that evidence, not the final word on the field.