Visual Memory Test

Visual short-term memory is the buffer holding what the screen looked like a moment ago. It is far smaller than intuition suggests: most adults reliably hold three to four objects, and that ceiling has survived decades of attempts to argue it higher.

It feels much larger because the world stays put. You experience a rich, detailed scene because you can always look again — the detail lives on the screen, not in your head. Change-detection tasks like this one remove that crutch, which is why catching a single changed unit in a twelve-unit field is startlingly hard the first time.

The practical consequence for a game: you are not holding the whole minimap. You are holding three or four things from it, and which three or four is a choice you are making, usually badly, under pressure.

The capacity ceiling is stubborn, but scores are not — because scores reflect strategy as much as capacity.

• Chunk the field into shapes. Four units in a rough diamond is one item, not four. Practised players group the field geometrically and store relationships rather than positions. This is the single biggest score lever.
• Take the whole field, not a corner. The instinct is to stare at one region and guarantee those units. It caps you at whatever fits in that region. Soft-focus the entire spread.
• Do not verbalise. Naming positions routes them through a slower verbal channel and costs you inside a sub-second window.
• Expect the big fields to feel impossible. They are above everyone’s ceiling — the weighting exists so that partial success on a hard field counts for more than certainty on an easy one.

Honest caveat: what improves with practice is your performance on this family of tasks. How far that transfers to noticing a missing jungler is genuinely uncertain, and anyone selling you certainty there is overselling.

Of all the measurements in the battery, short-term memory capacity is among the most obviously state-dependent — and unusually, it is one people notice themselves. The tired version of you does not feel slower; it feels like the field blanked.

Two things degrade together. Encoding gets worse: less makes it into the buffer during the brief exposure. And maintenance gets worse: what did make it in decays faster across the blank interval. A drop of a full item or more between rested and exhausted is unremarkable — which, against a ceiling of three or four, is a very large proportional loss.

This is the mechanism behind a specific, familiar experience: late-session games where you keep getting caught by things you would normally have registered. Not slower hands. A smaller buffer. See sleep deprivation and gaming performance for the wider picture.

Most “memory tests” online are recall tasks: watch a sequence, reproduce it. Those measure something real but different — our sequence memory test covers that ground, and order matters there.

Change detection measures capacity without order. Nothing is sequenced; you either had the unit in your buffer or you did not. That makes it a cleaner read on raw storage, and it is the paradigm labs reach for when they want to put a number on how many items a person can hold.

In game terms the split is intuitive. Sequence memory is a rotation or a path — order is the content. Change detection is a snapshot: who was on the map two seconds ago, no ordering required. Both are useful, and they come apart under fatigue in different ways.

Every minimap glance is a change-detection trial. You look for a fraction of a second, build a snapshot, look away, and act on that snapshot for the next several seconds. The whole skill is how much survives the glance.

Spatial working memory is also one of the clearer expert–novice separators in esports research samples — stronger players hold more of the map after the same glance. The honest read on causation is that it runs both ways and nobody has cleanly untangled it. What is not in doubt is that your own capacity shrinks when you are tired, which is a thing worth knowing before you queue rather than after. It is one slot of the daily check.