N-Back Training - Working Memory

N-back is the most studied working memory training task in cognitive science.

In N-back, you see a sequence of stimuli one at a time. For each item, decide whether it matches the one N steps back. At 1-back you compare each item to the previous; at 3-back you juggle three in mind.

The task directly loads the phonological and visuospatial buffers while demanding constant updating.

We score hits minus false alarms - guessing does not raise your score. Your history is saved locally.

Working memory is the mental workspace where you hold and manipulate information for seconds at a time. It is often discussed alongside fluid reasoning and academic task performance.

Jaeggi et al. (2008, PNAS) showed that dual N-back training improved fluid intelligence. Subsequent replications vary, but the task reliably improves working memory capacity.

N-back gauges how well you continuously refresh and rewrite working memory under load.

The task isolates the updating component of working memory: each new stimulus forces you to drop the oldest item, shift everything one position, and bind the newcomer to its place in the sequence. A strong result reflects an efficient, regularly maintained mental queue rather than raw storage size. Accuracy at higher N indexes how reliably that rolling update survives interference.

Unlike a simple span task, which asks you to hold and recall a fixed list, N-back never lets the contents settle; the target is a moving window, so it taxes monitoring and overwriting rather than retention alone. This continuous-updating demand is the feature that distinguishes it from digit span or Corsi block recall.

Stay at one level until matching feels automatic, then step up by exactly one.

Begin at 1-back to learn the rhythm, then move to 2-back, the standard working level. Only advance to 3-back once you can clear several runs with few false alarms. Focus on actively updating each item as it appears rather than waiting for a match, and use a steady inter-stimulus pace so you rehearse on a fixed cadence.

The common failure is passive watching: people wait to feel a match instead of refreshing the queue every single trial, which causes lag errors where you flag the wrong position. Resist subvocally chanting the whole list; that overloads the phonological loop. Treat non-matches as just as important as hits, since unflagged updates keep the window aligned.

Read hits minus false alarms together, never accuracy alone.

Your score is hits minus false alarms, so high accuracy means little if you also press on many non-matches. The N level you can sustain matters more than any single run: most people settle comfortably at 2-back, find 3-back effortful, and 4-back genuinely hard. Track whether your sustainable level and false-alarm rate trend in the right direction across sessions.

N-back is strongly state-dependent: fatigue, caffeine, stress, and time of day swing scores noticeably, so compare like with like. A good score reflects updating efficiency on this specific format, not general intelligence or memory in everyday life, and improvement here does not reliably transfer to unrelated cognitive tasks.

1. Watch the 3×3 grid. A square lights up every few seconds.
2. For each new position, ask: did this match the position N steps ago?
3. If it matches, press Match (or the Space key). Otherwise do nothing.
4. Accuracy matters more than speed.