A stream of items appears one at a time, and for each one you must decide whether it matches the item shown N steps earlier — answering yes or no on every single one. The twist is that every item you see is simultaneously a test of the past and new information you must carry forward, so the buffer you are checking against is never allowed to sit still.
Why airlines test it
Working-memory updating is distinct from simple retention: it is not enough to hold information, you must continuously overwrite it as new information arrives while still comparing against what came before. This is what the N-back format isolates — maintenance, manipulation, and interference control operating together under constant refresh. Selection boards weight this heavily because a pilot's mental picture of the aircraft's state is never static; it is being rewritten by ATC, systems, and the environment on a rolling basis, and a candidate who can hold information but cannot update it cleanly will fly on stale data without realising it.
In the cockpit
The running picture in your head — current cleared level, active runway, latest QNH, the last heading you were assigned — has to be updated the instant new information supersedes it, not merely added on top of what you already believed. The risk is not blank forgetting; it is holding onto a value that used to be correct and no longer is, because your attention was on the new item arriving rather than confirming it had actually overwritten the old one.
For example: on approach, ATC issues 'expect runway 27, QNH 1008' during initial descent, then twelve minutes later switches you with 'runway has changed, now landing runway 09, QNH 1006, contact Approach 119.7.' If the old runway and pressure setting are still sitting in your mental buffer alongside the new ones rather than having replaced them, you brief, configure, or set an altimeter for a runway and pressure that ATC withdrew several minutes earlier.
How to practise
Play ENAC M2-back numérique and hold the 2-back before you move to its 3-back mode — the jump in load is steep, and the useful skill is learning to let go of an item the instant it drops out of the window rather than gripping it. ENAC M2-back couleurs runs the same rule on unlabelled colour swatches, which adds a job the digits do for free: each colour has to be given a name of your own before it can be held. Visual N-Back, the form the DLR runs, loads both at once — a shape in a colour, where a match needs both to agree — and moves the goalposts by changing the lag from block to block. Pay attention to your false alarms as much as your misses: a false alarm means you responded to an item that felt familiar from the wrong lag, which is the N-back equivalent of acting on information that used to be true, so treat every false alarm as a rehearsal for catching stale data before you act on it. The Acoustic Memory test works the same capacity through your ears instead of your eyes: sequences are spoken, held and typed back, immediately or after a distraction-filled delay — the format the DLR's memory tests use — and the delayed mode in particular trains you to park information reliably while something else demands your attention. The Parameter Memory Test drills the COMPASS version of the same discipline: a panel of flight parameters — speed, radio, altitude, heading, fuel — shows briefly, the display goes dark, and you reproduce the values one row at a time, exactly as the real Memory subtest presents it. Grid Challenge recreates the dot-span format from Aon's smartPredict suite (officially gridChallenge): dots light up in order while symmetry, rotation and figure puzzles interrupt your rehearsal between each flash, and you click the positions back in their order at the end — exactly the interference under load this lesson is about.
The ENAC EPL selection tests this family with Memory Calculation — each calculation on screen for five seconds and a letter for one second after it, then the letters typed back in order into an empty row with nothing offered, over about twenty boards whose strings grow from three letters to nine. The ENAC Test Preparation course covers it in context. The Air France cadet pre-selection runs the plainest form of the same load: sixty-two items go past one at a time in ENAC M2-back numérique, and from the third one on, each asks for a yes or a no to the same question — is this the item from two places back? There is no third button, so an item you say nothing about is an item you got wrong. The same test has run on unlabelled colours in place of digits, and that form is ENAC M2-back couleurs. Almost nobody finds one item hard; what breaks is holding the chain to the end, and the errors come in pairs late in the run. Its later stage opens with span itself in ENAC Reversed Series: a cross marks the start, digits go past one at a time at a pace nothing you do can change, and you type the series back last digit first, from four digits up to nine with the length never announced. Then the same stage loads the rolling comparison with arithmetic and moves the goalposts, in ENAC Running Comparison: a hundred calculations in four runs of twenty-five, each on screen only long enough to be worked out, and what you compare the result against changes three times during the test — two calculations back, then the one just before, then a fixed value, then two back again, though the first three can come in another order. The arithmetic is easy on purpose; the errors cluster where the rule changes and the old one keeps running on its own. The Air France course sets each in its stage.