Dual n-back

The short answer

Dual n-back is a working-memory exercise. A square lights up on a 3 by 3 grid, and at the same time a voice says a letter.

On every turn you say whether the square is in the same place as n turns ago, and whether the letter is the same. It can be one, both or neither.

Most people start at 2-back, where you compare each turn with the one two turns earlier, and find it gets hard quickly.

It became famous in 2008, when a study suggested it could raise fluid intelligence: the kind of reasoning that IQ tests measure.

It is now one of the most studied brain-training tasks, and many people online call it the gold standard. Later studies have found mixed results, but the people who seem to gain most are the ones who keep improving and reach higher levels.

How the n-back task works

The n-back task was invented by Wayne Kirchner in 1958. He used it to study how memory for fast-changing information differs between younger and older adults.

Items appear one at a time. For each one, you decide whether it matches the item from n steps back.

At 1-back you compare with the item just before. At 2-back you skip one. At 3-back you skip two, and so on.

The hard part is that the target keeps moving, so you can't learn a list and relax. Every new item pushes the oldest one out, and you are always updating what you hold in mind.

Psychologists call this working-memory updating. The n-back is one of the standard lab tasks for measuring it.

What makes it "dual"

Dual n-back runs two streams at the same time. A square appears in one of nine places on a 3 by 3 grid, and a voice says a letter as it appears.

Each stream is checked on its own. On any turn the position might match, the letter might match, both might, or neither.

You press one button for a position match and another for a letter match.

Single, dual and triple n-back

How to play dual n-back

Here is the start of a 2-back round. Each row is one turn.

After each turn, you drop the oldest item and add the newest. A round usually runs for 20 to 30 turns.

How you are scored

Each stream has four possible outcomes on every turn.

Pressing on every turn would catch every match, but it would also rack up false alarms. Good scoring counts both.

Researchers usually use a measure called d′ (d-prime). It rewards telling matches from non-matches, not just pressing a lot.

How it gets harder

The main way is a bigger n. Going from 2-back to 3-back means holding one more turn in mind, and each step is a big jump.

Items can also come faster, which leaves less time to update.

Harder rounds add lures. A lure is an item that repeats one turn too early or too late, such as a letter that comes back after one turn in a 2-back round. It feels like a match, but it isn't one.

Beginner mistakes

Many beginners say the letters under their breath. That works at 2-back but stops working by 4-back, so try to hold the items without repeating them.

Guessing when you're unsure is another. False alarms count against you, so if you really don't know, leave it.

Most people also lean on the squares and let the letters slide. If your letter score is much lower than your position score, that's usually why.

What the research says

Some pages promise a higher IQ from dual n-back, and others say brain training never works. The studies are more mixed than either. Here they are in the order they came out.

The 2008 study that started it

Susanne Jaeggi and colleagues trained young adults on dual n-back that got harder as they improved. Groups trained for 8, 12, 17 or 19 days and were compared with people who did no training.

Scores on fluid-intelligence tests went up in the trained groups. The more days people trained, the bigger the gain.

The paper ran in PNAS and got wide press coverage. It was the first serious claim that a simple memory exercise could raise fluid intelligence.

The replications

Other labs tried to repeat it with stricter designs, and the results were more mixed.

Redick and colleagues (2013) gave young adults 20 sessions of dual n-back. One comparison group practised a visual search task that also got harder, and another did nothing.

The dual n-back group got much better at dual n-back. In this study, that didn't show up as higher scores on the ability tests.

Lawlor-Savage and Goghari (2016) ran five weeks of online training with 57 adults aged 30 to 60. The comparison group trained processing speed instead.

Here too, neither group moved much on overall measures of working memory, processing speed or fluid intelligence.

Why the comparison group matters

When the comparison group does nothing, the trained group gets extra time on computers, more contact with researchers and a reason to try hard on the retest.

A comparison group that does a different task, called an active control, levels those things out. Studies with one tend to find smaller effects, which is part of why researchers still disagree.

The meta-analyses

Single studies are noisy, so the clearest picture comes from pooling them.

Au and colleagues (2015) pooled 20 n-back training studies. They found a small but positive effect on fluid intelligence.

3 to 4 pointsThe average IQ gain Au et al. (2015) estimated across 20 n-back training studies.

Other researchers read the same evidence more cautiously.

Melby-Lervåg and Hulme (2016) reanalysed the same studies. They argued the effect came mostly from studies whose comparison group did nothing. With those set aside, the effect was smaller.

Melby-Lervåg, Redick and Hulme (2016) looked at working-memory training in general: 87 publications and 145 comparisons. Training reliably improved other working-memory tasks straight after training.

For far transfer, meaning skills a long way from the trained task such as reasoning, reading and arithmetic, the gains against an active control were much less clear.

Soveri and colleagues (2017) looked only at n-back training: 33 randomised controlled trials. Gains on untrained n-back tasks were medium-sized. Gains on other working-memory tasks, attention control and fluid intelligence were smaller, but still there.

So the biggest gains stay closest to the task you train.

Why your level matters

An average can hide big differences between people. Several studies found that the people who improve most on the task are the ones who gain most elsewhere.

In the 2008 study, the groups that trained for more days made bigger gains. In a 2011 study of school children, only the children who improved a lot on the training task got better on reasoning tests. Those gains were still there three months later.

Pahor, Seitz and Jaeggi (2022) ran three randomised trials. In all three, the people who improved most on a different n-back task also improved most on matrix reasoning, a common IQ test item.

That held even in the two trials where the training group as a whole didn't beat the comparison group.

One reading is that wider gains come through closer ones. You have to get better at memory tasks in general, which means pushing to harder levels, before anything wider shows up.

This fits what many people who train seriously say: the benefits come once you work your way up to higher n-back levels, not from staying at 2-back.

In January 2026, Meijer and colleagues gave 105 adults aged 62 to 93, all with memory or thinking complaints, either dual n-back or relational frame training. Both groups improved about the same amount on executive function.

Where that leaves things

We think dual n-back is worth doing for the skill itself: keeping track of things that keep changing.

No study can promise a set number of IQ points. The studies that did find gains point to the same advice: keep moving up rather than settling at a level that feels comfortable.

What happens in the brain

N-back is one of the most used tasks in brain-imaging studies of working memory.

A 2005 meta-analysis by Owen and colleagues pooled 24 imaging studies. The same areas turned up again and again, mostly in the front and towards the top and back of the brain (the prefrontal and parietal cortex).

That network is involved in holding information, updating it and choosing what to pay attention to. It is also active in many reasoning tasks.

That overlap is a big part of why researchers expected n-back training to help with reasoning. How much training changes the network itself is still being studied.

How it compares to other brain training

Dual n-back is often called the gold standard of brain training, mainly because it has been studied more than almost any other task. Relational frame training is the one it gets compared with most, since both have been studied for effects on fluid intelligence.

N-back trains how much you can hold and update from moment to moment. Relational training trains what you do with the information: combining it and drawing a conclusion. Our relational frame training guide covers that side.

For the bigger picture, our guide Does brain training work? sets n-back alongside the other types of training.

Dual n-back and the method of loci

The method of loci is a memory technique where you picture items placed along a route you know well. A 2021 study gave 148 young adults dual n-back, the method of loci or no training.

Both training groups improved on their own task and on digit span, a test of how many digits you can repeat back. Only the dual n-back group improved on a change-detection task.

The authors read that as broader transfer for dual n-back. The comparison group did no training, so the point above about comparison groups applies.

Is single n-back as good as dual?

In one direct comparison, Jaeggi and colleagues (2010) found single n-back produced similar gains to dual n-back.

If the audio stream doesn't suit you, one stream is a reasonable choice.

How to practise

The studies ran their training over weeks, not single sessions. These suggestions follow what they did, but there's no agreed best dose.

Use the studies as a guide

Most studies used sessions of 20 to 25 minutes on most days, for four to five weeks. That comes to about 20 sessions.

Shorter sessions are fine while you're getting used to the task.

Keep climbing

In the studies, the people who kept improving on the task were the ones who showed wider gains.

The training studies raised n when people did well and lowered it when they struggled. A level you have mastered stops being a workout, and a level far above you turns into guessing.

Aim to keep working your way up, from 2-back to 3-back and then 4-back and beyond.

Don't guess

If you really don't know whether a turn was a match, leave it. Guessing adds false alarms and doesn't train anything.

Check a task you don't train

If you care about wider gains, test yourself now and then on something you don't practise, such as a digit-span test or a different n-back task.

In the 2022 study above, the people who improved on a task they hadn't trained were the ones who also improved on reasoning.

How Portals does it

Portals is Neurobuff's dual n-back game. Underneath the visuals, it's the same task the studies used.

Nine portals stand in for the grid. A shape comes out of one portal for the position stream, and a voice reads a letter for the letter stream.

You can play single mode, with positions only, or dual mode with both streams. You choose the mode before a session starts.

Difficulty adapts to you. After each round, Portals works out your d′ and sets the next round from it. There's no level to pick.

In dual mode it goes by your weaker stream, so being strong on positions can't hide weak letters.

As you improve, turns come faster, from one every 2.5 seconds to one every 1.8 seconds, and lures start to appear. Then n goes up, all the way to 9-back.

Portals is one of five training games in Neurobuff. Start training to play it.

Frequently asked questions