Dual N-Back: The Most Researched Brain Training Task — What the Science Actually Says

In a dual n-back task you simultaneously track two streams: a spatial position (a square lighting up in a grid) and an auditory letter. Your job is to press a key whenever the current stimulus in either stream matches the one from n steps ago. At N=2, you compare now to two steps back; at N=3, three steps back. The dual aspect — visual and auditory together — is what makes the task demanding.

The task was developed in the 1950s by Wayne Kirchner as a laboratory measure of working memory load. It gained global attention in 2008 when Jaeggi et al. published a study claiming that training on dual n-back transferred to improvements in fluid intelligence — the ability to reason about new problems — as measured by Raven's Progressive Matrices.

The Jaeggi et al. (2008) finding triggered an enormous replication effort. Results were mixed. Some studies found transfer effects; others found none. A 2016 meta-analysis by Melby-Lervåg, Redick, and Hulme analysed 87 studies and concluded that working memory training does not reliably transfer to fluid intelligence or other cognitive abilities beyond the trained task itself.

The core problem is near versus far transfer. Dual n-back training reliably improves dual n-back performance (near transfer). But the evidence that this improvement carries over to untrained tasks like reasoning, reading comprehension, or academic performance (far transfer) is weak and inconsistent.

Setting aside the transfer debate, dual n-back places heavy demands on three well-established processes: working memory updating (continuously refreshing the n-step-back target), interference resolution (ignoring near-misses that match n-1 or n+1 steps ago), and divided attention (splitting focus between two independent streams). All three are clinically meaningful cognitive functions.

Consistent practitioners report that the task becomes effortful in a different way as n increases — less raw difficulty, more the specific sensation of actively holding and discarding information. This metacognitive awareness of your own memory operations is itself a useful cognitive skill.

Begin at N=1 (is the current stimulus the same as the previous one?) until accuracy exceeds 85% across several sessions. Move to N=2 only when N=1 feels automatic. Jumping ahead too quickly turns the task from a training stimulus into a frustrating failure spiral.

Sessions of 20–30 minutes, 4–5 days a week, are the protocol used in most published training studies. Beyond ~15 minutes of continuous n-back, error rates plateau and additional reps add fatigue rather than load. Quality over duration.