What Is the N-Back Test? How It Works and What It Shows

The n-back test is a working memory task. You see or hear a series of items, one at a time, and respond whenever the current item matches the one from n steps earlier, where n is usually 1, 2 or 3 (Owen et al., 2005). W. K. Kirchner introduced it in 1958 (Huang et al., 2025), and it has since become a standard working memory measure in cognitive neuroscience (Kane et al., 2007). Practice makes you better at the n-back itself. Whether those gains carry over to intelligence or other skills is still argued about, and the largest reviews find little evidence that they do (Melby-Lervåg et al., 2016; Soveri et al., 2017).

How the n-back works

Every level follows the same rule. You decide whether each item matches the one that appeared n items ago (Kane et al., 2007), and only the distance changes:

Level	Say “match” when the item is the same as…	Example sequence	Matches
1-back	the item right before it	T, K, K, R, R	the second K and the second R
2-back	the item two before it	T, K, T, R, K	the second T
3-back	the item three before it	T, K, R, T, M	the second T

The items can be letters, words, shapes, faces or pictures, shown on a screen or played as sounds. Some versions even use smells. Some versions ask whether the item itself repeats, and others ask whether it shows up in the same location (Owen et al., 2005).

The hard part is that the set you’re holding in mind keeps changing. Every new item has to be compared with the one n back and then added to the set while the oldest one drops off. Researchers describe the task as drawing on updating, maintenance and attentional control, with continuous updating as its defining feature (Huang et al., 2025). Near-misses add to the difficulty. These are “lure” items that repeat, but not from exactly n back. In 2-back and 3-back tests they triggered more false alarms than other non-matching items, because a recent item feels familiar even when it’s in the wrong spot (Kane et al., 2007).

Single vs. dual n-back

A single n-back has one stream of items. A dual n-back runs two streams at the same time. In the version used in a 2008 training study, a square appeared in one of eight places on a screen while a voice in headphones said one of eight consonants. Each pair lasted half a second, and a new one came every 3 seconds. Participants answered separately for location matches and letter matches. After each block, n went up by one if they made fewer than three errors in each stream and down by one if they made more than five (Jaeggi et al., 2008).

Where the n-back came from

Kirchner’s 1958 paper in the Journal of Experimental Psychology measured very short-term memory in younger and older adults. They watched a visual display with a rapidly moving light. Older participants’ performance fell off sooner, and they made more errors of omission and more random responses. Kirchner concluded they couldn’t organize incoming and outgoing information as quickly as younger people could (Kirchner, 1958).

What it measures

Working memory is usually described as a system for briefly holding information and working with it (Owen et al., 2005). The n-back looks like a working memory test, but it doesn’t line up neatly with other working memory tests. In one study, n-back scores were only weakly related to scores on an established working memory span task. That suggests the two don’t mainly measure the same thing (Kane et al., 2007).

How researchers use it

The n-back is a staple of brain imaging research. A 2005 meta-analysis pooled 24 n-back neuroimaging studies. It found consistent activity in frontal and parietal regions of the brain, including the dorsolateral and ventrolateral prefrontal cortex, across many versions of the task (Owen et al., 2005). Studies often compare 1-back, 2-back and 3-back so they can track what changes as the memory load goes up (same source). It’s also used as a training task. By 2017 there were enough published randomized controlled trials of n-back training in healthy adults (33 of them) for a meta-analysis (Soveri et al., 2017).

What the research says about n-back training

Training reliably improves n-back performance. What’s disputed is whether it improves anything else, and the evidence has shifted over time.

• 2008: Jaeggi and colleagues trained young adults on an adaptive dual n-back for 8 to 19 days. Compared with a group that got no training, the trained groups improved more on a test of fluid intelligence (reasoning about new problems), and the gains grew with more days of training. The authors noted that their study couldn’t show how long the gains would last (Jaeggi et al., 2008).
• 2013: Redick and colleagues randomly assigned young adults to 20 sessions of dual n-back, an active placebo (visual search training) or no contact. Both training groups got better at the task they practiced, but there was no positive transfer to any of the cognitive ability tests (Redick et al., 2013).
• 2015: A meta-analysis of 20 n-back studies in healthy adults found a small but significant positive effect on fluid intelligence (Au et al., 2015). Critics argued the result was misleading because it didn’t account for differences between groups at the start, or separate studies with active control groups from those without (Melby-Lervåg & Hulme, 2016). The original authors replied that the type of control group made no difference to the effect (Au et al., 2016).
• 2016: A larger meta-analysis of working memory training in general (87 publications, 145 comparisons) found reliable gains on working memory measures right after training. It found no convincing evidence of gains on “far” measures such as verbal and nonverbal ability, reading or arithmetic when trained groups were compared with control groups that did some other activity (Melby-Lervåg et al., 2016).
• 2017: The meta-analysis of n-back training trials mentioned above found a medium-sized effect on untrained n-back tasks. Effects on other working memory tasks, fluid intelligence and cognitive control were very small. The authors concluded that much of the benefit is specific to the task (Soveri et al., 2017).

When studies disagree like this, the newer and larger analyses deserve more weight. The 2016 and 2017 meta-analyses are both newer and larger than the 2015 one, and both find effects that stay mostly close to the trained task. A 2023 review of cognitive training more broadly concluded that the overall far-transfer effect is null (Gobet & Sala, 2023).

If you enjoy the n-back, practice will make you better at it, but the studies don’t show that it will make you smarter. For the bigger picture, see our guide on whether brain games improve memory.

Try a 2-back on paper

All you need is paper, a pen and a second sheet to cover the list.

1. Write a column of about 20 letters, using only five or six different letters so repeats come up often. Plant a few true 2-back matches (the same letter two lines apart).
2. Add a couple of lures: a letter that repeats right after itself, or three lines later.
3. Cover the column, then slide the cover down to show one letter at a time.
4. Starting with the third letter, say “match” whenever the letter you see is the same as the one two lines above it. Don’t peek back.
5. Uncover the list and check. Count your hits, the matches you missed and any false alarms.

To make it harder, ask a friend to read a fresh list aloud at a steady pace, or switch to 3-back.