MEMORY
GAME

Nine positions is more than almost anyone can hold. Ask healthy adults to watch a sequence tapped out on a board of nine blocks and repeat it back, one longer each time, and the average best run is about six. PsyToolkit's account of the standard version puts it at 6.2 blocks, with a standard deviation of 1.3, citing a study published in 2000.

That task has a name and a history. It is the Corsi block-tapping task, and its shape (nine positions, one showing, repeat the order, one more each level) has been a standard clinical measure of visuospatial memory span for more than fifty years. Because the sequence grows by exactly one a round, the last level you clear is your span. Your best level and your span are the same number.

The other thing with a history is the phrase "memory game" itself. It arrived from an unexpected direction.

Key Takeaways

• A nine-position board where you watch a sequence and repeat it back is the Corsi block-tapping task, devised by Philip Corsi in 1972 in doctoral work supervised by Brenda Milner.
• PsyToolkit's account of the task, citing Kessels and colleagues, puts the healthy adult average at a block span of 6.2 with a standard deviation of 1.3, and typical performance at five to seven against a theoretical maximum of nine.
• The British name for the older matching-pairs game is Pelmanism, traditionally credited to Christopher Louis Pelman and the Pelman Institute, founded in London in 1899.
• The Institute's own 1920 prospectus names a different founder, William J. Ennever, and carries a section headed "Not a Memory System".
• George Miller's "seven, plus or minus two" counted chunks, not items, and Miller himself called the recurring seven "a pernicious, Pythagorean coincidence".

Why is it called a memory game?

Because a Victorian correspondence school got there first, at least in Britain.

The older game that first carried the word is the matching-pairs one: cards face down in a grid, turn two over, keep them if they match. John McLeod's card-game reference records that it is "also known as Concentration or Memory", that "in Japan it is called Shinkei-suijaku", and that the British name Pelmanism comes "after Christopher Louis Pelman, who founded the 'Pelman Institute for the Scientific Development of Mind, Memory and Personality' in London in 1899".

So the British word for a memory game is a trading name, picked up from a business that sold mind training by post. Popular psychology outliving the evidence behind it is a recurring shape, and the personality test explainer traces another instance of it. And that attribution is not as tidy as it looks. Open the Pelman Institute of America's own 1920 prospectus, Mind and Memory: Pelmanism, and the masthead lists a president, a director of instruction, five heads of instructional staff, a librarian, a secretary and a four-man advisory board. The founder it names is "WILLIAM J. ENNEVER". Pelman appears nowhere on the page. The card-game tradition credits one man; the American arm's own paperwork credits another.

What Pelmanism was actually selling

Not memory, by 1920, and it says so out loud.

The prospectus contains a section headed "Not a Memory System", and is candid about the pivot:

Viewed historically, Pelmanism is a study in intelligent growth. Twenty-five years ago it was a simple memory training system. Time has broadened it.

What replaced it was a course in "the science of right thinking", promising that "mind-wandering will be cured and concentration will become a habit". The cover advertises "OVER 500,000 SUCCESSES IN ALL PARTS OF THE WORLD"; the inside cover lists offices in New York, London, Melbourne, Durban, Bombay and Toronto. This was a mass-market product sold by post, decades before anyone marketed a brain-training app.

There is a small dating puzzle in that quotation. "Twenty-five years ago", written in 1920, points at 1895, four years before the 1899 founding the card-game reference gives. Either the writer was rounding, or the memory system predates the institute that sold it.

The word stuck to the games anyway, which is how a card game and a nine-position sequence test came to share a name with a company that had publicly stopped selling memory training.

What is a memory span, and how is it measured?

A memory span is the longest sequence you can reproduce in the correct order. There are two standard kinds: a spatial one, where positions are tapped, and a verbal one, where a list is spoken. They do not give the same number.

Philip Corsi built the spatial one in 1972, as doctoral research supervised by Brenda Milner. Nine blocks sit in front of the participant. The experimenter taps a sequence; the participant taps it back in the same order; the sequence gets one longer. The block span, as PsyToolkit's description of the task puts it, "is defined as the longest sequence a participant can correctly repeat".

Roy Kessels and colleagues standardised the administration and published normative data in Applied Neuropsychology in 2000. They worked with seventy healthy participants and seventy patients with brain lesions. PsyToolkit's summary of that study gives the figure worth carrying: an average block span of 6.2, with typical healthy performance running between five and seven against a ceiling of nine.

What exactly the standardisation consisted of, meaning the trials at each sequence length and the stopping rule, sits behind the publisher's paywall. The definition used throughout here is the block span as PsyToolkit states it: the longest sequence correctly repeated.

This next part is standard background rather than a finding from the sources above. Verbal spans usually come out higher, which is why "seven" is the number most people quote. Digits are familiar, they can be said under the breath, and rehearsal helps in a way it does not help with a silent pattern of positions. That is a difference in material, not in brainpower.

Task	What it measures	What the figure is
Corsi block-tapping, nine positions	Visuospatial short-term span	A measured norm, about 6.2 blocks
Digit span, a spoken list repeated back	Verbal immediate memory span	Miller's constant was about seven chunks, not seven digits
Matching pairs, cards face down in a grid	Recognition and location recall	No standard span measure in the clinical literature

Two caveats before anyone treats 6.2 as a personal benchmark. It is a single mean for a sample of healthy adults, and span norms are ordinarily banded by age, so a global figure is a rough anchor rather than your target.

The second caveat is inference rather than citation. The sources describe the clinical apparatus as nine blocks placed irregularly rather than in a grid, but none of them tests what that does to scores. The reasoning is straightforward: an irregular scatter should resist being read as a shape, which is exactly the trick people use to beat their own span, so a tidy grid ought to be the easier board. On that reasoning, read 6.2 as a floor for grid versions rather than a like-for-like norm.

The clinical use also explains why the task is built the way it is. Kessels and colleagues report that right-hemisphere patients performed worse than left-hemisphere patients, and that the task "can be effectively used to assess visuospatial short-term memory in patients with brain damage, and is selective for the side of the lesion". The nine-position sequence is not an arbitrary puzzle shape. It is a diagnostic instrument that happens to be fun.

Where the nine-position ceiling actually comes from

The board shows you nine positions. Your span is not a property of the board.

Every position is visible the whole time, so the difficulty feels as though it ought to be about attention. But what you are storing is not the tiles. It is the order, and the order exists nowhere except in your head.

That is why the ceiling arrives so abruptly. Six is average, eight is genuinely unusual, and nine is the maximum of the clinical task rather than a normal result.

For the opposite case, where nothing has to be stored at all and the ceiling is nerve conduction, see the reflex tests.

Can you raise a memory span?

Not the raw capacity, on Miller's account. What you can change is what counts as one item.

George Miller made this the centre of his 1956 paper in Psychological Review, the source of the misquoted "seven, plus or minus two". Two things in it got lost almost immediately. First, he separated the span of absolute judgment from the span of immediate memory, calling them "quite different kinds of limitations". Second, his constant was not seven items but seven chunks: "the number of chunks of information is constant for immediate memory", where a chunk is however much material you have bundled into a single unit.

Miller was unconvinced by his own number:

Perhaps there is something deep and profound behind all these sevens, something just calling out for us to discover it. But I suspect that it is only a pernicious, Pythagorean coincidence.

His prescription was recoding: grouping raw input into fewer, larger units. That is worth understanding rather than just doing, because it explains why the familiar tricks work at all. Reading three positions as one diagonal does not give you a fourth slot. It puts three items into one of the slots you already had, and the slot does not notice the difference. A chunk is defined by meaning, not by size, which is why a phone number you know is one item and a phone number you do not is eleven.

That also sets the limit on the trick. Every grouping has to be one you can actually form in the second or two you are given, so the ceiling moves up by a level or two and then stops.

The ciphers are where recoding stops being a metaphor, and the quizzes test the opposite thing entirely: retrieval from long-term memory, where capacity is not the constraint.

Frequently asked questions

What is a good score on a nine-tile memory game?

Around six is average for a healthy adult, on the closest measured equivalent. PsyToolkit's account of the Corsi block-tapping task, citing Kessels and colleagues, gives a mean block span of 6.2 with a standard deviation of 1.3 and typical performance of five to seven blocks. Since the sequence grows by one a round, the last level you clear is your span. Clearing eight puts you well above the mean, and nine is the ceiling of the task rather than a target.

What is the Corsi block-tapping task?

A neuropsychological test of visuospatial short-term memory. Nine blocks are laid out, a sequence is tapped, and the participant repeats it in the same order, with the sequence growing by one each round. Philip Corsi devised it in 1972 during doctoral research supervised by Brenda Milner, and it remains a standard clinical measure, sensitive enough that performance differs by which hemisphere a brain lesion sits in.

Why is it called a memory game?

The word came into English game vocabulary through the matching-pairs card game, which is called Pelmanism in Britain after the Pelman Institute, a London correspondence school founded in 1899 that sold mind-training lessons by post. In the United States the same game is usually Concentration or Memory. The name later spread to sequence games of a completely different shape.

Is "seven, plus or minus two" the real limit?

It is misapplied more often than it is wrong. Miller's 1956 figure described about seven chunks, where a chunk is whatever you have managed to bundle into a single unit, and he was explicit that different spans are "quite different kinds of limitations". Spatial sequence spans measure lower, at around six, and unfamiliar material that resists grouping measures lower still.

Does practising a memory game improve your memory generally?

The honest answer, on Miller's own account of how anyone exceeds a span, is that what improves is your strategy rather than your capacity. His explanation for every case of someone beating the limit is recoding: you bundle items into larger units rather than acquiring more slots. Getting better at reading a nine-position board as shapes will make you better at nine-position boards. Treat any wider claim with suspicion until someone shows you the measurement.

Sequence span is only one kind of memory, and the rest of the arcade leans on the others. The daily puzzle runs on recall and habit; the reaction time test takes memory out of the loop altogether.

Sources

• Gijsbert Stoet, Corsi block-tapping task, PsyToolkit experiment library. Describes the nine-block task, its origin in Philip Corsi's 1972 doctoral work under Brenda Milner, the definition of block span, and the 6.2 figure attributed to Kessels and colleagues - retrieved 10 August 2026.
• Roy P. C. Kessels, Martine J. E. van Zandvoort, Albert Postma, L. Jaap Kappelle and Edward H. F. de Haan, The Corsi Block-Tapping Task: standardization and normative data, Applied Neuropsychology 7(4), 2000, 252-258. Citation and abstract read via PubMed; the publisher's full text is paywalled, which is why the 6.2 mean is credited above to PsyToolkit's summary - retrieved 10 August 2026.
• George A. Miller, The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information, Psychological Review 63, 1956, 81-97. Full text in Christopher Green's Classics in the History of Psychology archive at York University - retrieved 10 August 2026.
• Pelman Institute of America, Mind and Memory: Pelmanism, Over 500,000 Successes in All Parts of the World, New York, 1920. Read as the Internet Archive plain-text derivative; the officer list, the "Not a Memory System" section and the branch-office list are all from this edition - retrieved 10 August 2026.
• John McLeod, Pelmanism, pagat.com card-game rules. The attribution of the British name to Christopher Louis Pelman and the Pelman Institute, founded London 1899, and the alternative names Concentration, Memory and Shinkei-suijaku - retrieved 10 August 2026.