N-Back Training: What the Working Memory Research Actually Shows

In 2008, a study in the Proceedings of the National Academy of Sciences claimed something remarkable: a few weeks of a memory drill called the dual n-back task could raise fluid intelligence — the raw capacity to reason and solve novel problems, long considered one of the most stable traits a person has. Susanne Jaeggi and colleagues found a dose-response relationship, with more training days producing bigger gains. Brain-training apps built entire businesses on that single paper. The follow-up research, however, has been far less kind to the claim.

What N-Back Training Actually Is

The task sounds simple and is not. A stream of letters, positions, or sounds appears one at a time. In the "2-back" version, you press a button whenever the current item matches the one shown two steps earlier. Push the delay to three or four steps back and the task becomes brutal — you have to continuously update, and discard, a moving window of items held in working memory. Working memory itself is famously limited: George Miller's classic 1956 estimate put capacity at "seven plus or minus two" items, while Nelson Cowan's more rigorous 2001 review put the real number closer to four chunks when rehearsal strategies are controlled for. N-back training is, at its core, an attempt to push against that ceiling.

The Transfer Claim — and Why It Matters

Getting better at the n-back task itself is not interesting. Everyone gets better at a task they repeat hundreds of times. What made the Jaeggi study newsworthy was the claim of "far transfer" — that gains on this narrow drill carried over to fluid intelligence measured on completely different tests, like Raven's Progressive Matrices. If true, it would mean intelligence itself is more trainable than most cognitive scientists believed.

That is a big claim, and big claims get scrutinized. Zach Redick and colleagues ran a more tightly controlled replication in 2013, published in the Journal of Experimental Psychology: General, using an active control group that also completed a demanding but non-adaptive task. Participants who trained on the n-back task got substantially better at n-back — but showed no more improvement on fluid intelligence, working memory, or multitasking measures than the control group. The specific-skill gains were real. The transfer wasn't.

What the Meta-Analyses Actually Show

Single studies disagree, so what does the pooled evidence say? Jacky Au and colleagues ran a meta-analysis in 2015 in Psychonomic Bulletin & Review across 20 n-back studies and found a small but statistically significant effect on fluid intelligence — Hedges' g of roughly 0.24, a modest effect by any standard. Monica Melby-Lervåg and Charles Hulme took a harder line in their 2013 and 2016 meta-analyses, concluding that working memory training reliably improves performance on the trained task and closely related tasks, but the evidence for transfer to unrelated cognitive skills, academic achievement, or everyday functioning is weak to nonexistent once study quality and control-group design are accounted for.

The pattern across the literature is consistent: near transfer (to tasks that resemble n-back) is well supported. Far transfer (to general intelligence or unrelated skills) shows up in some analyses, disappears in others, and shrinks whenever the control condition is more rigorous. Reviewers who have tracked this field for years, including Anders Ericsson, have argued that most claimed cognitive transfer effects fail to hold up once you compare them against the kind of deliberate, domain-specific practice that actually builds expertise.

So Is It Worth Doing?

If your goal is a broad intelligence boost, the evidence does not support that outcome, and you should be skeptical of any app that promises it. If your goal is narrower — getting measurably better at holding and updating information under pressure for a specific task you care about, like mental math, note-taking during fast conversation, or certain working-memory-heavy job skills — near-transfer effects are real and replicated. The research on effective learning more broadly, summarized well in Make It Stick, points to the same underlying principle: durable cognitive gains come from effortful, spaced, task-specific practice rather than a single generic drill assumed to transfer everywhere.

How to Practice It Correctly

If you want to try it, an adaptive dual n-back task — one that raises or lowers the n-level based on your accuracy — produces the most consistent near-transfer gains in the literature. Sessions of 20 minutes, four to five days a week for at least four weeks, match the protocols used in the studies that found positive effects. Track your highest sustained n-level rather than raw accuracy; that number is what tends to climb over weeks of practice. Pair it with adequate sleep and, if you use a choline-based nootropic, know that citicoline supplementation has separate, modest evidence for supporting attention and working memory capacity independent of any training protocol.