N-Back Test
A square jumps around a 3×3 grid, one position every three seconds. Press Match whenever it lands where it was n steps ago. Unlike the other tests here, this one is not a ladder you fall off — it is a session you repeat, and it gets harder as you get better.
Watch the square move. Press Match when it is in the same place it was n steps ago — three blocks make a session.
Session complete
Your sensitivity this session: 0.00 d′ at 2-back
How this is scored. Most trials in an n-back run are not targets, so “percent correct” is a trap: press nothing at all and you still score around 72%, having caught nothing. The headline figure is d′ (d-prime) — the distance between how a target feels to you and how a non-target feels, in standard deviations, worked out from your hit rate and your false-alarm rate together. 0 means you cannot tell them apart. Around 1 is a real but shaky signal. Above 3 is a near-flawless session. Answering the same way on every trial scores exactly 0, whichever way it was.
Last 10 sessions
How the n-back test works
Each trial shows one lit square for half a second. Your job is to compare it with the square from n trials ago and press Match if they are in the same place. At 2-back you are holding two positions and updating both every three seconds; at 3-back, three. That constant discarding of the oldest item while a new one arrives is what makes n-back a working-memory task rather than a memory-span one — nothing is ever safely stored.
A session is three blocks of about twenty trials. The first n trials of each block have nothing behind them to match, so they are dealt but never scored — they show as hollow dots in the row beneath the grid.
Reading the row of dots
Each trial gets a dot. The outlined one is the trial you are on, which is worth watching: the square is only lit for half a second, so most of a trial is a dark grid, and the moving dot is how you can see the block is still running. Pressing Match marks that trial immediately, whether or not you were right, so a press is never silent.
By default the dots also settle into a verdict as each trial closes: a filled dot for a target you caught, a hollow ring for one you let past, a struck-through dot for a press on a trial that was not a target, and a faint dot for a non-target you correctly left alone. A running count sits under the row. If you would rather run a block with no hints — a target you missed is still information about the sequence — turn off Mark each trial right or wrong as I go and the dots stay neutral until the block ends, then resolve all at once.
The level moves on its own
Three clean blocks in a row — every target caught, no false alarms — and you go up a rung. Two missed targets in a single block and you drop one. Your rung is remembered between visits, so the task keeps sitting just past what you can comfortably do, which is the only setting at which this kind of practice is worth anything.
The level itself is deliberately not the number we chart. It is an integer that lives between 1 and 5 for essentially everyone, so a graph of it would be a flat line with the occasional step. It is a badge; the chart plots your d′, which moves continuously and can improve while your level stays put.
Today's boards are fixed
Each block is dealt from today's date together with the rung you are on, so reloading the page will not deal you an easier board — the next block is simply the next one. Tomorrow is a fresh set. Note that this does not mean everyone gets the same board: a target is defined relative to n, so a player at 2-back and a player at 4-back cannot meaningfully share a sequence.
What n-back training does and does not do
N-back is the most-studied working-memory training task there is, and the honest summary is that people reliably get better at n-back. Whether that improvement carries over to unrelated abilities is contested, and has been for over a decade — treat any site promising IQ points from a browser game with suspicion. What you can expect is a measurable, trackable improvement on this specific task, which is why the chart here plots your own sessions and nothing else.
Nothing you do here leaves your device: your level, streak and session history live in this browser's local storage, and there is no account and no server.
