N-back test: how many steps back can you hold?

Wayne Kirchner introduced the task in 1958 to study how memory for fast-changing information changes with age. It has since become one of the most used working memory tasks in brain-imaging research, because the load can be dialled up one step at a time while everything else stays the same.

Each level asks for more mental bookkeeping, not more memory span. At 2-back you hold two positions, compare the new one with the older of them, then shift everything along by one.

This is the question that made the task famous, and the evidence has moved a long way since the headline.

• 2008: the claim. Susanne Jaeggi and colleagues trained adults on dual n-back and reported gains on a fluid intelligence test that grew with the number of training sessions. It was published in PNAS and widely reported.
• 2013: a placebo-controlled test. Thomas Redick and colleagues gave one group 20 sessions of adaptive dual n-back, another an adaptive visual search task as a placebo, and a third nothing. Both trained groups got better at their own task. Neither improved on any of the tests of fluid intelligence, working memory, multitasking or knowledge.
• 2015 and 2016: the meta-analyses. A 2015 meta-analysis (Au and colleagues) found a small average gain, about 3 to 4 IQ points. A 2016 meta-analysis of 145 comparisons (Melby-Lervåg, Redick and Hulme) found transfer only in studies whose control group did nothing, not in those with an active control, and concluded that working memory training does not improve intelligence. The two teams still disagree about how to code control groups, which is why you will see both numbers quoted.

The fair summary: practise n-back because it is a hard, satisfying test of focus, and you will get better at it. Do not expect it to raise your IQ. Our evidence review of what affects IQ grades brain training next to education, sleep and nutrition.

Working memory is one of the four index scores on the WAIS-IV and one of five on the WAIS-5 and WISC-V. It is measured with tasks such as repeating digit strings forwards, backwards and in order, not with the n-back. In fact the n-back correlates only weakly with those standard span tasks (Kane and colleagues, 2007), so a strong n-back result does not translate neatly into a working memory index.

If you want the closest thing to the IQ-test version, try the digit span test. To see where working memory sits in a full profile, read WAIS index scores.