Memory
Visual Memory Test
A grid appears with half its squares lit, and you get three seconds to take it in. Then it clears and you mark what was there, in any order you like. Every pattern you nail adds another square. Your score is the biggest one you got exactly right, printed alongside the odds you simply got lucky.
Visual memory
A grid appears with half its squares lit, and you get three seconds to take it in. When it clears, mark the ones that were lit. Order doesn't matter here, and you can change your mind as much as you like before you submit. Every pattern you nail adds another square; miss twice at the same size and that's your run.
What your equipment contributed
- Your display
- not measuredassuming a typical 60 Hz panel
- Pattern shown for
- 3000 ±16.7 msquantised to whole frames
- Matrix rendered at
- not measuredCSS pixels wide
Timing barely reaches this score. The three-second exposure is quantised to whole frames by your display, so it is really 3000 ±16.7 ms — one part in 180, against a result that is a count of cells rather than a duration. Your screen is the limit that matters here, and we can only half state it. The task this derives from is a printed card at a normal reading distance, so every participant sees a pattern of roughly the same physical size. Yours is whatever your viewport allows across — and a CSS pixel is not a fixed physical length, so we cannot convert that to centimetres, and nothing in a web page can measure how far away you are sitting. We report the number the browser gives us and decline to derive a physical size from it.
How this is measured
How the pattern is presented
The whole matrix appears at once and stays for 3000 ms, then clears. Nothing is presented item by item, and nothing is timed about your answer — there is no clock running while you mark cells. The exposure begins and ends inside a requestAnimationFrame callback, so it starts on a frame your display can actually present and lasts a whole number of frames rather than an arbitrary slice of one. During the exposure the cells are disabled, so an early tap cannot be recorded at all.
Why exactly half the cells are filled
It is the defining feature of the published task, and it does real work. If the proportion filled were free to vary, a level could be quietly easier or harder than the level number printed beside it — three cells scattered in a large grid is a different problem from three in a small one. Holding density at half means the only thing that changes as you climb is size. It also makes the guessing odds exact, which is the one number this page can offer in place of a reference range.
What guessing could get you
The answer is always “choose exactly half the cells”, so the number of possible answers at a level of k filled cells is exactly C(2k, k) — 6 at the opening size, 184,756 at ten cells, 155,117,520 at fifteen. That is arithmetic, not an estimate, and it is printed with your result. The practical reading: luck can buy you the smallest score on the ladder and essentially nothing above it, because a size is only reached by passing the one below it.
How the test ends, and what the score is
Patterns start at 2 filled cells and grow by one every time you reproduce one exactly. Two misses at the same size end the test — one slip is not enough to establish that a size is beyond you, which is why the published span procedures give more than one attempt at each level. Your score is the largest pattern you reproduced exactly, counted in filled cells, which is the convention the source task uses. A near miss scores nothing: 14 of 15 cells is not a partial span, and we show you the near miss rather than folding it into the number.
The ceiling is the ladder, not you
The source task's largest pattern is 15 filled cells in a 5×6 matrix, and ours stops in the same place. Reproduce it and we report ≥15, not 15 — past that point we stopped measuring you and started measuring the ladder. Three sizes on the way up have no near-square matrix available at all: 7, 11 and 13 filled cells need 14, 22 and 26 cells, which only divide as 7×2, 11×2 and 13×2. We show the strip rather than break the half-filled rule, and a strip is a slightly different visual problem from a block. It is stated here because it is a real seam in the instrument.
What we cannot tell
We cannot tell whether you remembered a shape or described one. Small patterns can be put into words — “top row, then two down the left” — and once they are, this stops being a visual measurement and becomes a verbal one. The problem is well enough known that a paper exists about limiting it, cited below. Our patterns are random rather than screened for describability, so we cannot claim to have controlled it; we can only tell you it is there. We also cannot tell whether you traced the pattern with a finger, sketched it, or were interrupted. And because the task is purely visual, we deliberately do not announce the filled cells to a screen reader: doing so would convert it into a verbal list task and produce a number that looks like this one but measures something else.
How your result compares
It doesn't, and this is the first page on this site where that is the honest answer. Every other test here either cites a published range and shows it beside your score, or would if one existed. Here one does exist and we still won't show it, for reasons worth stating plainly rather than burying.
Why no reference range is shown
We have not read the norms. The task's normative data was published in the printed test manual (Thames Valley Test Company, 1997), not in an open paper. Quoting a mean we have taken from a secondary summary would be exactly the move this site exists to refuse — the whole point of a citation is that you can go and check it.
And it would not transfer if we had. The published task uses two fixed parallel sets of standardised patterns, printed on cards, shown by an examiner at a normal reading distance, and reproduced with a pencil on paper. Ours are drawn at random every round, on a display of unknown physical size at an unknown distance, marked with a fingertip or a mouse. Random patterns are not interchangeable with screened ones — some random layouts are far more describable than others, which is the verbal-coding problem above.
The precedent is already on this site. Our sequence memory page cites a pilot study that ran the same spatial task on a physical board and on a tablet and found they scored differently. If moving one task to a screen changes it measurably, assuming this one survived the move unchanged would be a guess dressed as a comparison.
What you get instead is the guessing odds — 1 in 6 at the opening size, 1 in 155,117,520 at the ceiling. It is not a percentile and does not pretend to be. It answers a narrower question exactly: how much of your score could have come from luck. On this page that is the only comparison figure that is actually true.
And no percentile from our own visitors, ever. We could compute one tomorrow — it would be self-selected, unvetted, taken on screens of every size, and it would look exactly as authoritative as the ones our competitors show. That is precisely the objection.
The task this derives from: Della Sala, S., Gray, C., Baddeley, A., Allamano, N., & Wilson, L. (1999). Pattern span: a tool for unwelding visuo-spatial memory. Neuropsychologia, 37(10), 1189–1199. (cited for the procedure — the three-second exposure, the half-filled matrices from 2×2 to 5×6, and scoring by the filled cells in the largest pattern recalled — not for a range.)
On verbal coding of visual patterns: Brown, L. A., Forbes, D., & McConnell, J. (2006). Limiting the use of verbal coding in the Visual Patterns Test. The Quarterly Journal of Experimental Psychology, 59(7), 1169–1176. (cited for the existence of the problem and of work to limit it; we have not applied its method.)
Last reviewed: July 2026
Frequently asked questions
Why is there no percentile or comparison figure on this page?
Because we could not find one we are able to verify. Norms for this task exist, but they were published in the printed test manual (Thames Valley Test Company, 1997) rather than in an open paper — we have not read them, and we will not quote figures we have not read. Even if we had, they were collected on paper with a fixed set of standardised patterns, administered face to face; ours are generated at random on a screen of unknown size. Every other test on this site either cites a range or says none applies. This is the first one where the honest answer is the second.
Could I pass a level by guessing?
At the opening size, yes — a two-cell pattern in a 2×2 matrix has only six possible answers, so a blind guess is right one time in 6. It stops mattering almost immediately, because the answer is always "choose exactly half the cells" and the number of ways to do that grows fast: 1 in 184,756 at ten cells, 1 in 155,117,520 at fifteen. Since you only reach a size by passing the one before it, guessing can buy you the smallest score on the ladder and essentially nothing beyond it. The odds for your own final level are printed with your result.
Why doesn't the order I mark the cells in matter?
Because order is what the other test measures. Our sequence memory test lights blocks one at a time and asks for them back in order, which is a spatial and sequential task. This one was designed to strip that component out: the whole pattern arrives at once, and what is left to remember is its shape. Marking the cells in a particular order would smuggle the sequential part back in and make the two tests measure the same thing twice.
Why can I change my mind here when the sequence test won't let me?
Because the two source tasks differ. The block-tapping task is administered by an examiner watching taps, with no undo, so our version makes a tap final. This one is administered with a pencil on a blank printed grid, where reconsidering before you hand the sheet back is ordinary. It also matters practically: the largest matrix has 26 cells, small ones on a phone, and a fingertip landing one cell over is not a memory event. Making it correctable removes a source of error that has nothing to do with what is being measured.