N-back levels explained: N=1, N=2, N=3 and beyond

What the level actually means

N is the number of trials between the current item and the item you compare it with. At N=1, compare with the previous trial; at N=2, compare with two trials ago; at N=3, compare with three trials ago. N counts trials, not seconds. In dual n-back, apply that same distance independently to the position and the sound. A position match does not imply a sound match, and you may need to answer either channel, both, or neither.

Why changing N changes the task

Increasing N extends the interval across which you must keep track of the sequence while new items arrive. It does not change the comparison rule. Difficulty also depends on the pace, the number of channels, the similarity of the items, the response window and how errors are scored. Two apps showing N=3 can therefore present different challenges. When comparing your sessions, keep those settings as similar as possible. N alone is not a standardized score of memory capacity.

N=1: learn the two decisions

At N=1, a sound or position is a match when it repeats the immediately previous sound or position. This can be a useful place to learn the controls, distinguish the two response channels and understand when not to press. It can also be challenging depending on the pace, the person and the settings. There is no universal number of introductory sessions, and staying at N=1 is not inherently useless. Move on when you understand the rules and want to try a longer comparison interval.

N=2: compare with two trials ago

Consider the spoken sequence A, B, A. At the third trial, the current A matches the A from two trials earlier, so the sound response is appropriate at N=2. The middle B is not the comparison item. For position, compare the third position with the first position independently. In the sequence A, A, B, the second A repeats a recent item, but that immediate repetition is not an N=2 sound match. Familiarity is not enough: the distance must also be right.

N=3: keep the distance precise

At N=3, the fourth item is compared with the first, the fifth with the second, and so on. With A, B, C, A, the fourth sound is a match. With A, B, A, C, the third A is not an N=3 match: it appeared too early to have a comparison three trials back. In the usual task, the first N trials establish the sequence before a full N-back comparison is possible. Follow the instructions of the particular implementation you are using.

N=4 and N=5: longer intervals, not intelligence labels

At N=4, compare the fifth trial with the first. At N=5, compare the sixth with the first. These are longer comparison intervals; they do not automatically make a session better. There is no population percentile attached to these numbers on this site, and a high N is not an IQ score. If you reach a higher level, record the settings and the pattern of errors as well as the level itself. A single best block is a personal result, not a reliable general assessment.

Single n-back and dual n-back are different tasks

Single n-back asks for one comparison stream, such as position. Dual n-back asks for two streams at once, commonly position and sound. You must keep the responses separate as well as track the sequence. There is no fixed conversion saying that dual N=2 equals single N=3. If you switch modes, treat the results as different series in your notes. A change in score after adding a stream does not by itself establish a loss or gain in general cognitive ability.

What adaptive difficulty does

An adaptive implementation changes difficulty according to a rule based on recent performance. Depending on the implementation, that rule may use errors, accuracy or separate channel scores, and may operate after a block or a session. It can help avoid remaining at a setting that is consistently too easy or too difficult for that task. The exact threshold is part of the app design, not a universal research prescription. Moving down a level is a settings adjustment, not a judgment about you.

Read the errors as well as the level

Separate a hit, a missed match, a false alarm and a correct non-response. Pressing on every trial can detect many matches while producing many false alarms. Never pressing can avoid false alarms while missing every match. A total percentage is hard to interpret unless you know how it combines those outcomes. Where the app provides them, compare the two channels and their errors. If the display only provides an aggregate score, avoid inventing more detailed conclusions than it supports.

When should I try N=3 after N=2?

First check that you can explain the N=2 comparison and distinguish the two response streams. Trying N=3 means comparing with three turns earlier; lower accuracy at first does not diagnose a loss of ability. In this app, access can depend both on progression and on Pro features. A progression unlock and a purchase restriction are different; consult the settings guide and the labels in your installed version.

When to stay or move down

Stay at the same level, slow down if the app permits it, or move down when you repeatedly lose the rule or can no longer tell why a response was wrong. You can also pause. A short reset may make it easier to distinguish a controls problem from a sequence-tracking problem. Do not force a higher N simply to match another person’s result. Their settings, experience and scoring may differ, and a shared level number does not make the sessions equivalent.

How to interpret a plateau

A flat level trend has several possible explanations: ordinary variation, changed settings, inconsistent practice conditions or a genuine plateau on the task. The graph alone cannot tell you which explanation is correct. Compare several sessions rather than just the best score; note interruptions and changes in pace or mode. There is no guarantee that a plateau will resolve on its own or on a fixed schedule. Continuing, simplifying or taking a break are choices about practice, not tests of willpower.

A progression you can actually evaluate

Begin with a level at which you can follow the rule, record the mode and settings, and review a few comparable sessions. Change one setting at a time when possible. Keep the conclusion specific: improved accuracy at a particular N shows improvement on that task under those conditions. It does not establish an increase in IQ, a clinical benefit or better performance at school or work. The separate evidence guide explains why improvement on the practiced task and transfer to other activities are different questions.

Further reading

• Kane et al. (2007), Working memory, attention control, and the n-back task, Journal of Experimental Psychology: Learning, Memory, and Cognition
• Jaeggi, Buschkuehl, Jonides & Perrig (2008), Improving fluid intelligence with training on working memory, PNAS
• Cowan (2010), The Magical Mystery Four: How is working memory capacity limited, and why?, Current Directions in Psychological Science