How long and how often to train dual n-back: what the studies actually did

Why a schedule matters more than a session

Almost everyone who tries dual n-back has a good first week and then stops. The reason is rarely the difficulty of the task. It is that no decision was ever made about when it would happen, so it competes with everything else each day and eventually loses. A schedule you can keep is worth more than an ideal schedule you abandon, and that principle should shape every choice below.

What the original study did

In the 2008 study by Jaeggi and colleagues, participants trained on a dual n-back task for between eight and nineteen daily sessions, with each session lasting roughly twenty-five minutes. The groups that trained longer showed larger changes on the outcome measure the authors used. That design is where most of the popular advice about daily half-hour sessions comes from, even though the study was not designed to identify an optimal schedule.

What later research found

The picture became more complicated. Redick and colleagues in 2013 used an active control group and found no improvement in fluid intelligence after n-back training. Meta-analyses since have disagreed with each other, and Soveri and colleagues in 2017 found clear improvement on n-back tasks themselves, some on closely related tasks, and little beyond. In practice that means no study can tell you the schedule that will make you smarter, because that outcome has not been reliably demonstrated in the first place.

What a schedule can honestly promise

It can make you better at the exercise, and that improvement is real, measurable and reasonably quick to appear. It can give you a number that becomes harder to raise, which is a legitimate reason to practise anything. What it cannot promise is a change in your memory at work, your grades or your reasoning, and any schedule sold on that basis is being sold on evidence that does not exist.

Short and frequent beats long and occasional

The most robust finding about practice scheduling in the wider learning literature is the spacing effect: the same total amount of practice produces better retention when it is spread over time rather than concentrated. Cepeda and colleagues reviewed this across a large body of studies in 2006 and examined the timing question directly in 2008. For dual n-back the practical translation is straightforward. Fifteen minutes on five days beats seventy-five minutes on one.

A schedule you can keep

A reasonable starting point is ten to fifteen minutes, four or five days a week, at a time of day you can defend. That is short enough to survive a bad week and long enough to produce a stable measurement. Attach it to something that already happens, such as the coffee you make anyway, rather than to an intention. If you miss a day, the schedule has not failed; if you miss a week, look at when you placed it rather than at your motivation.

Why longer sessions backfire

Accuracy in this task falls with fatigue, so a forty-minute session typically contains a good first half and a second half that drags your average down and rehearses sloppy decisions. You end up with a worse measurement and a worse habit. If you genuinely want more practice, add a second short session later in the day rather than extending the first. Two separate blocks also give you the spacing benefit for free.

Time of day, and why it matters more than you think

This task is unusually sensitive to alertness. Sleep, caffeine and time of day move scores by an amount that can easily exceed several weeks of real improvement. That has two consequences. Pick a slot and keep it, so your sessions are comparable. And do not read anything into a session you did at midnight after a long day, other than that you were tired.

Let the level move both ways

Most implementations raise the level when you do well and lower it when you do not. Resist the urge to lock the level to protect a number. A level you have outgrown produces high scores that measure nothing, and a level you cannot hold produces guessing. Bouncing between two levels every few blocks is not a failure. It is the system finding the edge where the exercise actually does something.

How to structure a single session

Keep it simple. Two or three blocks at your current level, with a short pause between them, and no changes to the settings during the session. Do not warm up at an easier level, because that adds a variable without adding information. Stop while the last block is still as accurate as the first, which is usually earlier than you expect.

What to change, and how often

Change one thing at a time and leave it for at least two weeks. Speed, number of trials, number of streams and level are four separate variables, and changing more than one at a time means your new numbers cannot be compared with your old ones. Most people who feel stuck have actually been changing settings, so their record is a pile of unrelated results rather than a series.

Rest days and gaps

A missed day costs almost nothing. A missed fortnight costs some of the task familiarity, which comes back within a session or two. There is no evidence for a specific optimal number of rest days in this exercise, so choose a rhythm you will keep rather than one you have read about. If you are training every day and dreading it, four days a week that you actually do is the better schedule.

Reading progress over weeks, not days

Judge a fortnight, not a session. Session-to-session variation in this task is large enough that a single bad result means very little, and a single excellent one means no more. Look at whether your accuracy at a fixed level is rising and steadying across five or more sessions, and treat the highest level you have ever touched as a trophy rather than a measurement.

When to stop or change course

Stop if you are guessing to keep a level, because the record has stopped meaning anything. Change course if you have been flat for two months at the same accuracy under stable conditions, and consider whether your sessions are too long or badly timed before concluding you have hit a ceiling. And stop entirely if you are only continuing because you believe it will make you smarter, since that is the one outcome the research does not support.

A schedule to start from

Week one: N=1 for two sessions to learn the controls, then N=2. Weeks two to four: ten to fifteen minutes, four or five days a week, same time of day, level adaptive, settings untouched. From week five: review your accuracy across the whole month before changing anything. Dual N-Back shows your session results and practice history, so the review takes a minute and needs no spreadsheet.

Further reading

• Jaeggi, Buschkuehl, Jonides & Perrig (2008), Improving fluid intelligence with training on working memory, PNAS
• Melby-Lervåg & Hulme (2013), Is working memory training effective? A meta-analytic review, Developmental Psychology
• Au et al. (2015), Improving fluid intelligence with training on working memory: a meta-analysis, Psychonomic Bulletin & Review
• Redick et al. (2013), No evidence of intelligence improvement after working memory training, Journal of Experimental Psychology: General
• Melby-Lervåg, Redick & Hulme (2016), Working memory training does not improve performance on measures of intelligence or other measures of “far transfer”, Perspectives on Psychological Science
• Cepeda et al. (2006), Distributed practice in verbal recall tasks: a review and quantitative synthesis, Psychological Bulletin
• Cepeda et al. (2008), Spacing effects in learning: a temporal ridgeline of optimal retention, Psychological Science
• Soveri et al. (2017), Working memory training revisited: a multi-level meta-analysis of n-back training studies, Psychonomic Bulletin & Review