How many things can you hold in visual memory?

Visual working memory is the short-lived store that keeps a picture of what you just saw, so you can compare it, act on it or point back to it. It's what lets you notice that something on a page has moved, or tap a pattern back a second after it disappears.

Researchers treat it as separate from memory for words and numbers. Steven Luck and Edward Vogel open their best-known paper by noting that short-term memory "can be divided into separate subsystems for verbal information and visual information". That's why holding a pattern of squares can feel quite different from holding a phone number.

In a study published in Nature in 1997, Luck and Vogel measured how many simple features, such as colors or the tilt of a line, people could hold in mind at once. Their answer: information about only four colors or orientations at a time.

The surprise came next. People could hold both the color and the orientation of four objects just as well as one feature of each. With objects that combined four features, they kept track of sixteen features in all, as long as those features belonged to four objects. The limit, the authors concluded, is on whole objects, not on individual details.

Later work has broadly agreed. A 2011 review by Timothy Brady, Talia Konkle and George Alvarez describes the classic results as a model in which "a fixed number of objects (3–4) were remembered". It is the same "about four" that turns up in research on memory for digits.

Number isn't the only thing that matters. The 2011 review reports broad agreement that visual working memory can hold "1 item with greater fidelity than 3–4 items". Its authors describe capacity as quantity times fidelity: how many things you hold, and how precisely.

That trade-off is part of why our visual memory game is possible past four squares. You only need to know which squares lit up, not their exact shade or size, so each one costs less to hold.

Four objects is not a hard ceiling on what you can see and recall. The same review gathers evidence that structure lets people hold more. When items can be grouped, "such grouping facilitates visual working memory, allowing more shapes to be remembered".

Regularities help too. In one study by the same authors, people who learned which colors tended to appear together came to remember nearly twice as many colors, about five or six, as a comparison group.

Knowledge works the same way. The review describes the original chunking research as "using stored knowledge to recode stimuli into a new format to increase capacity", the visual cousin of turning a string of digits into a year or a time.

Where things are matters as well. Items are "encoded along with spatial context information", and the review cites work showing that changing the layout around an item makes changes harder to spot. In other words, you remember positions partly as a layout, not as a list of separate points.

People also keep a sense of the group as a whole, such as its overall spread or average. The review concludes that these "ensemble statistics appear to be integrated with individual item representations": your memory of the overall pattern shapes your memory of each square in it.

The review's own summary: how much you can hold "depends on what kind of items you are trying to remember", how they are presented, and your history with them. A score in a memory game is a measure of all three, not only of capacity.

Remembering where things were is one task. Remembering the order they came in is another. The classic test of the second is the block-tapping task, named after Philip Corsi, whose 1972 doctoral research at McGill University used it.

In the usual version, a person watches someone tap a series of identical blocks laid out on a board, then taps the same blocks in the same order. The series grows until they can no longer repeat it. Like digit span, it measures a span, but a span of positions in order rather than of numbers.

The task has been used for decades, but not always in the same way. A 1998 review by Berch, Krikorian and Huha found considerable inconsistency in how it was given and scored, and even in the board itself. In 2000, Roy Kessels and colleagues published a standard way of running and scoring it. The lesson for any version, including ours, is that the exact setup changes the numbers.

Our visual memory game asks where. Our sequence memory game asks where and in what order.

• Visual memory starts with 3 lit squares on a 3 by 3 grid. Each level adds a square, the grid grows every three levels up to 7 by 7, and the flash shortens from 1.1 seconds toward 0.7. You can tap the squares back in any order, and you have three lives.
• Sequence memory uses a fixed 3 by 3 grid, closer to a block-tapping board. Level 1 lights one tile, and every level adds one more to the sequence. A tile can light more than once, and one mistake ends the run.
• By level 2 of visual memory you are holding four squares, right at the research's "about four". The first band, First Glance, runs to level 4, which is six squares. By level 5 you are holding seven. Going further depends on seeing the squares as a few shapes rather than many separate points.

We sort results into bands rather than ranking you against other players, because we don't collect the data that would make a ranking honest.

• Look for shapes in the lit squares: a line, a corner, a block, a diagonal. A shape is one thing to remember instead of several.
• Take in the whole grid at once rather than hunting square by square while the flash is on.
• In sequence memory, split a long sequence into short runs of two or three taps, and notice when the path doubles back or repeats a tile.
• Keep your eyes on the grid during the playback, and wait until it ends before you start tapping.
• Compare scores on the same device, since the grid looks quite different on a phone and on a laptop.

These are games. A score on one try, on one device, shows how you did in that moment. It isn't a measure of your memory in general, and it says nothing about your health.