N-Back vs Dual N-Back: What They Train and Whether It Transfers

If you have spent any time on cognitive-training forums, you have run into N-back. It is the king of brain-training tasks for one reason: a 2008 paper claimed that practising the harder "dual N-back" version increased fluid intelligence, which would be the holy grail of cognitive training if true. Since then, the field has spent fifteen years arguing about whether the original finding holds.

The short answer: single N-back trains a specific kind of working memory and is hard but tractable. Dual N-back is harder, demands more of the brain, and may produce slightly larger gains, but the broad claim that it raises general intelligence is, at best, weakly supported.

If you want to test your basic working memory without N-back specifically, our Number Memory test and Sequence Memory test measure similar capacities in a more conventional format.

What N-back actually is

The N-back task presents a sequence of stimuli, one at a time. After each stimulus, you press a button if the current one matches the one from N steps ago.

For example, 2-back with letters:

H, K, L, K, J, L, K, K

You would press the button on:

• The fourth letter (K) because it matches K from two letters ago.
• Possibly on the seventh letter (K), since the second-back letter is L (no match), so actually no.

You can see why it gets confusing. As N grows, the demand on working memory grows roughly linearly. Most adults can manage 2-back without much difficulty. 3-back is hard. 4-back is very hard. 5-back is brutal.

The task forces continuous updating of working memory. Each new item must be encoded, the oldest item must be dropped, and the comparison must happen, all within a couple of seconds before the next item arrives.

How dual N-back differs

Dual N-back runs two N-back streams simultaneously. The most common setup is:

• A visual stream: squares light up in different positions on a 3x3 grid.
• An auditory stream: letters are spoken aloud, one at a time.

You have to track both streams independently. You press one key when the current square position matches the position from N steps ago, and a different key when the current letter matches the letter from N steps ago.

You can imagine why this is harder. You are essentially running two N-back tasks at once, and your working memory has to maintain two separate queues without confusing them.

People typically start dual 2-back at around 60 to 70 percent accuracy. With practice, they progress to 3-back, 4-back, and beyond. Highly trained practitioners can reach dual 7-back or 8-back, which is genuinely extraordinary.

The Jaeggi study and what it claimed

Susanne Jaeggi and colleagues' 2008 paper in PNAS was the spark. They had healthy young adults train on dual N-back for about 20 minutes per day for 8, 12, 17, or 19 sessions, then tested their fluid intelligence using a standard reasoning test.

The headline finding: the more days of training, the larger the improvement in fluid intelligence, with the longest-trained group showing the largest gains.

This was a stunning claim. Fluid intelligence had long been thought of as largely fixed in adulthood, a kind of cognitive bedrock. If 20 minutes a day of a specific game could move it, that would change a lot of assumptions.

The paper sparked an immediate industry of N-back apps and an immediate scientific debate about whether the finding would replicate.

What replication has shown

The replication record is mixed at best.

Redick and colleagues' 2013 study was an early major attempt. They used a more rigorous design with active control groups (people doing other cognitively demanding tasks rather than passive controls). They did not find the fluid intelligence benefit. Multiple subsequent replications have come back negative.

The 2015 Au meta-analysis by Jaeggi's collaborators argued that across the available studies, dual N-back training did produce a small but real improvement in fluid intelligence (effect size around 0.24, considered small in psychology). They acknowledged that the effect was smaller than the original paper suggested but defended its reality.

The 2017 Soveri meta-analysis by independent researchers reached a more skeptical conclusion. They found large improvements on the trained task, modest improvements on similar working memory tasks, and no reliable improvement on fluid intelligence when active control groups were used.

A reasonable summary: the original strong claim of large fluid intelligence gains is not supported. There may be a small effect under specific conditions, but it is smaller and less reliable than the marketing for these apps suggests.

What N-back definitely does train

Setting aside the controversial transfer claims, the more modest claims about N-back have good support.

The trained task itself. Your N-back score climbs substantially with practice. Reaching dual 4-back from a starting point of 2-back is a typical journey over 6 to 12 weeks of regular training.

Closely related working memory tasks. Some transfer to other tests of the same kind of attention-and-updating working memory has been documented.

A particular kind of attentional discipline. Sustained attention under high cognitive load. Many practitioners report feeling sharper at multitasking that involves rapid switching between streams of information.

Frustration tolerance. Honestly, the most underrated benefit. Dual N-back is hard. Practising it teaches you to sit with sustained mental difficulty without quitting. That habit transfers in ways the cognitive science does not measure.

Single vs dual N-back, which to choose

If you have decided you want to train N-back despite the modest evidence for transfer, here is how to choose.

Single N-back is a good starting point for beginners. It trains the same updating-and-comparison mechanism without the dual-stream complexity. Most people can reach 5-back or 6-back with consistent practice in a few weeks. The task is challenging but not overwhelming.

Dual N-back is the more demanding option and the one with most of the published transfer research behind it. It requires more focus, more energy, and tolerates fewer distractions. A 25-minute session leaves most people genuinely mentally tired.

For practical training, most experienced practitioners recommend starting with single N-back for a week or two to get familiar with the format, then switching to dual N-back if you want the higher-intensity version.

A realistic training schedule

If you want to give N-back a fair trial:

• 20 to 25 minutes per session.
• 5 to 6 sessions per week.
• For at least 4 weeks before evaluating whether it is doing anything.

This is what most of the positive studies have used. Shorter or less frequent training produces less.

Within each session, most apps use an adaptive algorithm. If you do well at 2-back, the next round becomes 3-back. If you struggle, it drops back to 2-back. This keeps you near the edge of your capacity, which is where training adaptations tend to happen.

Track your N-level over time. Most beginners start at 2-back and reach 3-back within a week or two. The jump from 3-back to 4-back often takes several weeks. 5-back and beyond is many months of work.

Do not be discouraged by plateaus. They are normal. Many practitioners stall at 4-back for weeks before suddenly breaking through.

What to do if N-back is not for you

It is okay to not like N-back. The task is unusually demanding and unusually unpleasant. Many people who try it for two weeks decide the experience is not worth the modest evidence for benefit.

If you want cognitive engagement without the N-back format, several alternatives provide similar working-memory pressure in friendlier packaging:

• Strategic games (chess, go, certain board games) train similar capacities with much higher inherent reward.
• Musical instrument practice trains working memory and many other cognitive functions in a meaningfully transferable way.
• Language learning is one of the most evidence-supported cognitive activities for sustained brain engagement.

None of these have the laboratory cleanliness of N-back, but they are easier to sustain over years, which probably matters more for cumulative cognitive benefit than the specific exercise format.

The honest summary

N-back trains a specific kind of working memory task and does so reliably. Whether that training produces broader cognitive benefits is contested. The original strong claim is weakly supported. The modest claim, that you get better at N-back and at similar working-memory tasks, is well supported.

If you enjoy the challenge of N-back, it is harmless and modestly beneficial. If you do not enjoy it, do something else cognitively challenging instead. The boring evidence-based ways to improve cognitive function (sleep, exercise, learning hard skills, managing stress) outperform any single training game by wide margins, as we cover in our working memory training piece.

FAQ

For other cognitive training games and tests, see our brain games hub. For the broader evidence on whether brain training transfers, see working memory training evidence.