N-Back Test — Free Working Memory Test Online
Watch cells light up on a 3×3 grid and press Match whenever the position repeats from exactly 2 steps back. 30 fixed-pace trials (28 scored, exactly 9 true matches) measure your working-memory balanced accuracy.
What this test measures
Spatial working memory — the ability to hold a stream of positions in mind, update it continuously, and compare the newest item against one seen earlier. A 3×3 grid lights up one cell at a time — each for a fixed 1.5 seconds followed by a 0.4-second blank gap — over 30 trials. Answering locks in your response but never shortens the trial, so every position gets equal encoding time. From the third trial onward you answer every step: press Match (M) if the current cell is the same as the one from exactly 2 steps back, No Match (N) otherwise — staying silent counts as a No Match answer. Exactly 9 of the 28 scored trials are true matches. The first 2 trials are observation only (no 2-back reference exists yet). Your score is balanced accuracy: the average of your hit rate on the 9 true matches and your correct-rejection rate on the 19 non-matches — the signal-detection scoring used in laboratory n-back studies. Random guessing, or never answering at all, sits at the 50% chance level.
The science behind it
The n-back task was introduced by Kirchner in 1958 and became the workhorse working-memory paradigm of cognitive neuroscience: Owen et al. 2005 (Human Brain Mapping) meta-analyzed dozens of neuroimaging studies and showed it reliably engages the fronto-parietal working memory network. Its fame outside the lab comes from Jaeggi et al. 2008 (PNAS), which reported that dual n-back training improved fluid intelligence and launched a brain-training boom. Replications were less kind: Redick et al. 2013 (Journal of Experimental Psychology: General), a randomized placebo-controlled study, found no intelligence gains, and Melby-Lervåg & Hulme's 2013 meta-analysis concluded working-memory training yields mainly short-lived, task-specific effects. The task remains an excellent measure of working-memory updating — what is disputed is transfer, not measurement.
How to improve your working memory
N-back performance itself is highly trainable: with practice, hit rates climb and false alarms fall, pushing balanced accuracy from the untrained average of about 75% into the mid-90s (Jaeggi et al. 2010). Useful strategies: rehearse the last two positions as a rolling pair, and commit to an answer on every trial — with only 9 matches against 19 non-matches, a missed match costs your score more than a stray false alarm. Watch out for lures: repeats at the wrong distance (3 steps back instead of 2) cause the most errors. Be honest about what improves, though: practice reliably raises your n-back score and performance on similar memory tasks, while far transfer to intelligence has repeatedly failed to replicate (Redick et al. 2013; Melby-Lervåg & Hulme 2013). Sleep and sustained attention matter too — working memory is among the first abilities to degrade when you are tired.
Frequently Asked Questions
What is a good n-back score?
Scores here are balanced accuracy — the average of your hit rate on true matches and your correct-rejection rate on non-matches — so random guessing (or never answering) sits at exactly 50%. Untrained adults land around 75% with an SD of roughly 12% (estimated from the behavioral 2-back literature, Jaeggi et al. 2010, pending recalibration on live data). About 87% puts you around the top 16%, and 95%+ is roughly the top 5%. Practiced players push into the mid-90s, so compare a first attempt against untrained norms.
Is this a dual n-back test?
No — this is a single spatial n-back: one visual stream of positions. The dual n-back made famous by Jaeggi et al. 2008 adds a simultaneous auditory letter stream, so you track two sequences at once. Single n-back is the standard laboratory form (in both verbal and spatial variants — Owen et al. 2005); this version uses the spatial one.
Does n-back training increase IQ?
Probably not. Jaeggi et al. 2008 (PNAS) reported fluid intelligence gains after dual n-back training, but a randomized placebo-controlled replication found no intelligence improvement (Redick et al. 2013), and the Melby-Lervåg & Hulme 2013 meta-analysis found mostly short-lived, task-specific gains. Training will raise your n-back score — claims beyond that are not well supported.
What does 2-back mean?
The N in n-back is how many steps back you compare against. In 2-back you judge whether the current position matches the one from two trials earlier, which forces continuous updating: every new item must be stored while an older one is compared and discarded. 1-back is much easier, 3-back much harder; 2-back is the standard research difficulty.
Why don't the first two trials count?
A 2-back judgment needs a position from two steps earlier, and that reference doesn't exist on trials 1 and 2. Those trials are observation only — you just memorize them — and scoring starts on trial 3. That is why 30 trials produce 28 scored answers.
More Cognitive Tests
Watch the squares flash, then repeat the sequence. Each success adds one step — how long a sequence can you hold?
Judge digits under two alternating rules — blue asks 'greater than 5?', orange asks 'is it even?' — with the rule switching every 2 trials. Your score is your switch cost in milliseconds.
Track a bouncing dot and click it the moment it turns red — while classifying digits as odd or even. 16 dual-task rounds measure how much accuracy you keep when both tasks compete.