← Intel

CHUNK IT — HOW MEMORY GAMES ARE BEATEN

The wrong way to remember a sequence

Sequence Memory lights tiles one at a time and asks you to click them back in the same order. Each round adds one more tile, so the sequence grows until you miss.

The instinct is to hold the whole sequence as a string of positions — tile, tile, tile, tile. That works for about five or six items and then it collapses, because working memory holds a handful of items, not a long list.

The fix is not a better memory. It is fewer items.

Groups, not items

A sequence of nine tiles becomes three groups of three if you say them as three words. The limit is on the number of things you are holding, so shorter groups do not make the job slightly easier — they change what the job is.

Two rules make grouping work in a game that is already moving:

  • Name as it lights, not after. Give each tile a short tag — a position, a colour, a letter — while it flashes. Naming afterwards is recall; naming during is encoding.
  • Group at a natural boundary. Three is a good size. Four works if the groups are visually distinct. Do not group by arbitrary halves of the sequence.

Recognition is not recall

Did You See It asks a different question: has this word already appeared in this stage? That is recognition, and it is much cheaper than recall — you do not need to reconstruct the list, only to notice that a word feels familiar.

The mistake is answering as if it were a memory test. Hesitating to search your memory for each word costs two seconds per wrong call and a lot more per hesitation. Trust the feeling of familiarity; it is usually right, and the game is built to reward it.

Colours are words, not pixels

Chroma Catch shows a colour for five seconds and then hides it. You rebuild it with hue, saturation and brightness sliders, and the gap is scored as a perceptual distance.

A remembered pixel decays. A remembered word does not. "Dusty teal, low saturation, medium brightness" survives the five seconds and gives you three numbers to aim at. The same idea as chunking: convert a hard-to-hold value into something language-shaped.

Counting under a clock

Color Count flashes a grid and then asks how many cells were a given colour. Counting one cell at a time is the losing strategy.

Count in rows or in pairs — the same reduction, applied to quantity instead of order. "Two rows of three plus one" is four things to remember instead of seven. When the grid disappears, that description is still available; a raw total usually is not.

The common thread

Every game on this site that tests memory is really testing how you encode, not how much you can hold. Groups instead of items, words instead of pixels, structure instead of raw counts.

That is why the difficulty rises the way it does. Sequence Memory adds one tile per round, which is trivial to hold — until you stop grouping. The games are not getting longer than your memory. They are waiting for you to find a smaller way to hold them.