What this test measures
A set of squares lights up simultaneously, holds for a moment, and vanishes. Your job is to click every square that was lit, in any order you like. Clear the pattern and the next level adds one more square; the grid itself grows from 3×3 to 6×6 as you climb. A wrong click costs one of your three lives — hard mode gives you just one, and keeps its own record — and clicking a square you have already found costs nothing. Your score is the last level you completed, and the result screen leads with how many tiles that level held at once.
What you are using is short-term picture memory: the power to hold a layout in your head for a few seconds with nothing to look at. The moment the squares go dark, the only copy of the pattern is the one in your head. Everything now depends on how well that copy survives the few seconds before you start clicking. Psychologists call the amount you can hold your span. The striking thing about span is how small it is next to what the eye takes in. Your visual system handles millions of pixels at once. What survives the moment the squares vanish is a handful of items.
This is not the same as Sequence Memory
The two tests look similar and measure different things, which is exactly why both are here. In Sequence Memory the squares light one after another and you must reproduce the order. Order is the whole task. Knowing which four squares were involved earns you nothing if you tap them in the wrong sequence. That is order memory. It uses the same machinery that lets you repeat a phone number back.
Here everything appears at once and order is irrelevant. There is no sequence to rehearse, so the strategies that carry you through a sequence test do not transfer. Instead you encode a shape: a diagonal, a cluster in one corner, three along the top edge. People often score quite differently on the two. A large gap between them isn't a contradiction, it's information. It usually means one of the two encoding strategies is doing most of your work.
Why there is no percentile here
Most sites would tell you that your result beats some percentage of players. We don't. No published normative distribution matches this task, and inventing one would make the number worse than useless.
There is a well-established clinical relative: the Corsi block-tapping task, in which an examiner touches wooden blocks and the patient reproduces the pattern. Normative data exists for it — Kessels and colleagues (2000) published standardisation data for exactly that purpose. But the Corsi task is sequential, administered by a person, with physical blocks. Grading a self-paced browser game with simultaneous stimuli against those tables means comparing two different tasks and calling the difference a score.
The research that does speak to simultaneous patterns points somewhere interesting. Luck and Vogel (1997) found that picture memory holds roughly four things at once. Cowan (2001) looked across many studies and argued the same: about four chunks, once repeating and grouping are ruled out. Those experiments used brief, precisely timed displays under laboratory conditions. This game shows the pattern for about one to two seconds on whatever screen you happen to be holding. If most people reach level eight or ten here — well past four squares — that is not a refutation of the research. It is evidence that you are grouping, which is the next section.
Why you can beat the four-item limit
The limit is counted in chunks, not squares. A chunk is whatever your brain has learned to treat as one thing. Three squares in a row are one chunk if you see them as a line. Four squares at the corners of a rectangle are one chunk if you see the rectangle. This is why players who see shapes climb higher than players who memorise single positions. They spend one slot where the other person spends four.
Practical versions of the same trick:
Look for lines and edges first. Rows, columns and diagonals are already familiar objects and cost almost nothing to store. Then split the grid into quadrants and count within each. "Two top-left, three bottom-right" is two chunks and a pair of small numbers, which is far cheaper than six coordinates. Giving the shape a silent name, an "L", an "arrow", a "staircase", adds word memory on top of picture memory, and two weak copies of a pattern beat one.
On dense boards, remember the gaps instead. When most of the grid is lit, the unlit squares are the smaller set, and memorising four holes is easier than memorising fourteen tiles.
Two habits protect what you have stored. Keep your eyes still during the display, because sweeping around costs time and fragments the layout, while a steady gaze near the centre lets peripheral vision take the whole board as one image. And click the squares you are sure about first. The pattern decays while you deliberate, so spending your confidence early leaves fewer squares to reconstruct from a fading trace.
What moves your score between runs
Expect variation. Span measures are noisy in individuals, and a single run tells you less than the shape of five runs. Screen size matters more than people expect. On a phone, the whole 6×6 board fits in one glance, which helps. On a big monitor, your eyes have to travel, and that breaks the picture into pieces. Interruptions are brutal here. A notification during the flash does worse than distract you. It overwrites you: whatever your eyes take in next steals the slots your pattern was using.
Tiredness fails in a specific way. You still see the pattern clearly — then lose it in the second or two before your first click. That gap is where maintenance happens, and maintenance is the first thing tiredness takes.