Dual N-Back · Guides
Ten dual n-back mistakes that quietly stall your progress
The habits that make dual n-back scores look better than they are, from guessing and level-chasing to training tired, and what to do instead.
Guessing instead of deciding
The single most common mistake is pressing when you are not sure, because pressing sometimes pays off and doing nothing never does. Over a session this inflates your hit rate and your level while destroying the meaning of both. The fix is not to become timid. It is to notice the difference between a decision and a reflex, and to let the misses happen. A session where you miss several real matches but press only on decisions tells you something. A session where you pressed your way to a good score tells you nothing you can use next week.
Chasing the level instead of the accuracy
A level is a single number and it is very tempting. But the level you reach is a consequence of your accuracy, not a substitute for it, and it can be reached by luck or by pressing often. Two people at the same level can be performing very differently. Judge a week of training by whether your accuracy at a given level is rising and steadying, not by the highest N you have touched. The level will follow, and when it does it will mean something.
Rehearsing out loud
Beginners often try to keep the sequence alive by repeating it, silently or under their breath. It works at N=1 and it fails from N=2 upward, because maintaining the rehearsal competes for the same attention the updating needs. Worse, it creates a second task to fail at. Most people who break through a plateau describe letting go of rehearsal rather than perfecting it, and answering on a feeling of recognition instead. That feeling is unreliable at first and becomes less so with practice, which is the part that actually improves.
Treating both streams as one
The position stream and the sound stream are separate questions with separate answers. Confusing them produces a very specific error: you know something matched, you press, and it was the other stream. If this is happening a lot, slow down and deliberately answer one stream first for a few blocks before returning to both. The goal is not to split your attention evenly, which is not really possible, but to keep the two answers from bleeding into each other.
Training for too long in one sitting
Long sessions feel virtuous and they corrupt your data. Accuracy in this task falls with fatigue, so a forty-minute session usually contains a good first half and a second half that drags your average down and teaches you sloppy habits. Short and frequent beats long and occasional, both for the measurement and for whether you keep doing it. If you want more practice, add a second short session later in the day rather than extending the first.
Ignoring false alarms
Most people look at how many matches they caught. Far fewer look at how often they pressed when there was no match at all. That second number is where guessing hides. If your false alarms are climbing while your hits stay flat, you are not improving, you are pressing more. Any honest review of your own progress has to look at both, and a session with fewer hits and far fewer false alarms is usually the better one.
Changing the settings every session
Speed, number of trials, number of streams, level. Change one and your new scores are not comparable with your old ones. Change three and you have no series at all, just a pile of unrelated numbers. Pick a configuration, keep it for a couple of weeks, and change one thing at a time when you do change something. This is the difference between practising and fiddling, and it is also what makes a plateau readable instead of mysterious.
Only training when you feel sharp
Training exclusively on your good days produces a flattering record and a fragile skill. It also removes the sessions that would tell you how much of your score depends on being well rested. You do not have to train when you are exhausted, but a normal week should include ordinary days. The alternative is a graph that looks like steady progress and is really a graph of how you felt.
Reading a single bad session as evidence
Session-to-session variation in this task is large enough that one bad result means very little on its own. People routinely drop a level for a day and conclude the method has stopped working, or that they have lost ground. Give any conclusion at least four or five sessions to earn itself. The corollary is equally true and less welcome: one excellent session is not evidence either, and the level you touched once is not the level you are at.
Expecting the wrong thing to improve
If you start dual n-back expecting your memory in daily life to change, you will spend the whole time looking for signs of something the research does not reliably support, and you will find them whether or not they are there. The companion guide on the evidence goes through this in detail. Expect to get better at the exercise, expect that to be measurable, and treat anything else as a bonus you cannot claim.
Practising without a fixed pace
Some implementations let you slow the interval down. Slowing it makes the task easier in a way that does not show up in the level, so a level reached at a slow pace is not comparable with the same level at the standard pace. If you use a slower setting to learn the controls, note it, and return to a fixed pace before you start treating your numbers as a series.
Comparing yourself to strangers
Forum posts about N=6 and N=7 are not a benchmark. There is no verification, no standard configuration, and no way to know what pace, how many streams or how much guessing produced those numbers. The only comparison that means anything in this exercise is with your own recent sessions under the same settings. Dual N-Back shows your accuracy per session for exactly that reason.
A short checklist
Before your next session, settle four things and leave them alone for two weeks: the level, the pace, the number of streams and the session length. During the session, press on decisions and let the misses go. Afterwards, write down your accuracy and your false alarms, not just the level. Then judge nothing until you have five sessions to look at. That is most of what separates people who improve from people who accumulate numbers.
Questions
Why is my n-back level going up but my accuracy is not?
Almost always because you are pressing more often rather than remembering better. Guessing raises your hit rate and your level while raising false alarms at the same time. Look at how often you press when there was no match before concluding you have improved.
Should I use a strategy or memory technique?
Explicit rehearsal works at N=1 and becomes a liability from N=2 upward, because maintaining it competes with the updating the task requires. Most people who break a plateau describe letting go of strategy and answering on recognition instead.
How long should a dual n-back session be?
Short enough that the second half is as accurate as the first. Long sessions produce a good first half and a fatigued second half that drags your average down and teaches sloppy habits. Two short sessions beat one long one.
Is it bad to train tired?
Training only when you feel sharp produces a flattering record and hides how much your score depends on rest. Include ordinary days, but do not draw conclusions from a session where you were exhausted.