What Is the N-Back Task? How It Works and What It Trains

Here is the task in one sentence: watch a stream of items go by, and for each one, say whether it matches the item you saw n steps earlier. There is no puzzle and no trick — just a memory load you keep updating, item after item, without ever setting it down.

In the common version, a square lights up somewhere on a 3×3 grid, one flash every second or two. On a 2-back, every flash asks you a yes/no question: is this the same position as the flash before last? The stream does not pause while you decide. By the time you have answered, the next flash has already replaced part of what you were holding — which is the point. The task is built so that standing still is impossible.

What the n measures

The n is how far back you look. A 1-back asks whether each item repeats the one right before it, which is easy to the point of boredom. The 2-back is the laboratory standard. The 3-back is where most people start to crack: you are holding three positions, every new item demands a comparison against the oldest one, and the two newer ones still have to survive for their own comparisons later.

To manage any of this you have to update, drop, and reorder continuously — close to the textbook definition of working memory, the part of the mind that holds information while you operate on it. That is why the n-back became the default working-memory task in brain-imaging research: the load scales smoothly with n, so one task can be tuned from easy to brutal for anyone.

Where the task came from

The n-back was born at 3M, not a university. In 1958, Wayne K. Kirchner, a psychologist in the company's personnel research group, published “Age differences in short-term retention of rapidly changing information” in the Journal of Experimental Psychology (DOI 10.1037/h0043688). His question was practical: how does the ability to hold fast-changing information hold up across a working lifetime? His instrument was a task where the answer was always n items back — too far back for a glance, never far enough to relax.

Kirchner found what he expected: the age gap widened as n grew. Older workers kept pace on the 1-back and fell progressively behind on the 2-back and 3-back. The paper has since been cited thousands of times, and the task built to answer a personnel question now runs inside brain scanners.

The dual n-back and the 2008 claim

For fifty years the n-back stayed a specialist's tool. Then in 2008 it went mainstream. Susanne Jaeggi and colleagues reported in PNAS that training on a deliberately punishing version — the dual n-back, which makes you track a spoken letter and a lit grid position at the same time — raised scores on tests of fluid intelligence, with gains that grew the more days people trained (PNAS, 2008).

Fluid intelligence is reasoning about problems you have never seen before, as opposed to applying what you already know, and it had spent decades looking fixed. If a few weeks of a game could move it, the implications ran far past the laboratory. The paper became one of the most cited studies of its decade and launched an industry: most brain-training apps of the 2010s were, one way or another, selling the Jaeggi result.

What happened when others tried it

The replication record is rougher than the headline. Meta-analyses by Melby-Lervåg and Hulme pooled the training studies and kept finding the same pattern: people get measurably better at the task they practiced and its close relatives, but gains on distant measures — intelligence tests, attention, school achievement — shrink to almost nothing once the studies use real control groups (Perspectives on Psychological Science, 2016). The jargon is near transfer versus far transfer. Near is real. Far, so far, mostly is not.

That is roughly where the field sits today: some analyses still find a small positive effect of n-back training on fluid reasoning, others find none, and no serious paper defends the large gains of the original study. Meanwhile the task itself was never in dispute — it measures working memory just fine. It is the upgrade promise that outran the evidence.

How we use it

Our N-Back game is the classic single version: a 3×3 grid, one lit position per turn, and you flag every match with the position two turns back. Twenty scoring turns, accuracy and timing saved locally so you can watch your own baseline move. Play N-Back.

The honest pitch: practicing n-back makes you better at n-back, quickly and measurably. Whether it makes you better at anything else is a live scientific argument, and we are not going to pretend this site settles it.

Frequently asked questions