Sequence Memory Test — Repeat the Growing Pattern

How to play

Squares light up in order, then get tapped back in the same order. Tap to pause.

Watch the orderSquares in a 3×3 grid light up one at a time, each with its step number.
Click them backRepeat the squares in exactly the order they lit.
It growsEach level you clear adds one more step.
One mistake ends itYour score is the last level you completed.

What this test measures

When you hold a phone number in your head, you use verbal working memory. This game uses its spatial partner, visuospatial working memory. It is a short-term store for where things happened and in what order.

You use it when you watch someone tap a door code and try to repeat it. You also use it when you glance at a map and keep the next few turns in mind.

Every level asks for two things at once: which squares lit up, and in what order. If you get every position right but swap two steps, you still fail. That is why each extra step feels so much heavier than the last.

What counts as a good score

This is the only test on the site that shows a percentile. We compare you with 340 adults from Facchin and colleagues (2024). Their average forward Corsi span was 4.97 (standard deviation 1.03). The badge on your result follows this table.

Level completed	Badge	What it means
14 or more	Superhuman	Nine standard deviations above the mean. Nobody in the study came close.
10-13	Pro tier	Well past the highest score in the study, which was 8.
7-9	Skilled	Two standard deviations above average. That puts you among the best of the 340.
5-6	Standard	The adult average. Level 6 is one step above it.
4 or fewer	Needs practice	A little under average. It usually comes up after a run or two.

Our grid and step numbers make this version a bit easier than the lab task. The average also covers ages 21 to 89. See "How our version differs" below.

Real scores recorded on this site

So far, 275 runs of this test have been recorded here. Computers and phones score quite differently, so each gets its own table.

Computer (mouse)

166 runs · median level 7

Range	Share
14 or more · Superhuman	5%
10-13 · Pro tier	20%
7-9 · Skilled	34%
5-6 · Standard	16%
4 or fewer · Needs practice	24%

Phone (touch)

109 runs · median level 7

Range	Share
14 or more · Superhuman	5%
10-13 · Pro tier	22%
7-9 · Skilled	30%
5-6 · Standard	28%
4 or fewer · Needs practice	16%

Every finished run counts, so someone who plays five times adds five runs. The numbers are recounted every day, and a device table shows up once it passes 30 runs. Last updated: 1 Nov 2026.

The science

This game comes straight from the Corsi block-tapping task. Philip Corsi invented it in 1972 for his doctoral research at McGill University. A tester taps some of nine wooden blocks fixed to a board, and the person being tested taps the same order back.

The longest sequence someone can repeat is their Corsi span. It became one of the standard measures of visuospatial working memory. Over decades of use, the task split into many lab versions, so in 2000 Roy Kessels and colleagues published a standard version.

The percentile after each run comes from Facchin and colleagues (2024), who tested 340 healthy Italian adults aged 21 to 89. The mean span was 4.97 with a standard deviation of 1.03, and scores ranged from 2 to 8.

How our version differs from the lab task

Three differences matter when you read your result.

The lab task scatters its blocks unevenly. We use a neat 3×3 grid where every square has an obvious name like "top-left", so verbal memory can help.
In the lab, an examiner taps by hand. Here, in normal mode, each square lights for half a second (700 ms if your device asks for reduced motion). You answer at your own pace with nobody watching.
Every lit square shows its step number along with the color change. That keeps the pattern readable for people who cannot tell the colors apart. It also gives you something to repeat in words.

All three make the screen version a little easier than the original. So read the percentile as a reference point against a published dataset, not as a clinical score.

How to remember longer sequences

Around level 5 or 6, plain memorizing stops working for most people. From there, strategy decides how far you get.

Chunk into shapes

Three squares that form an L, a diagonal or a row are one shape, not three places. A nine-step sequence becomes three shapes.

Trace a path

Picture a line travelling from square to square as the pattern plays. One continuous route is easier to replay than a list of points.

Name the squares

Whisper "one top-left, two center" as they light. When the spatial picture starts to fade, verbal memory backs it up.

Keep your eyes on the grid

A glance at a notification wipes the spatial trace fast. Keep your eyes on the middle of the grid until the pattern ends.

When your score climbs with practice, most of the gain is better strategy, not bigger memory. We still count that as skill. But a practiced level 8 and a first-try level 8 are different achievements.

Frequently asked questions

Why did the pattern not play?

The tab was almost certainly hidden. Browsers slow their clocks in background tabs to save power, which would scramble the timing. So the game pauses instead of playing a garbled sequence. Bring the tab to the front and press Start, and the same level plays again from the beginning.

Can I change how fast the squares light up?

Normal mode lights each square for 500 ms. Hard mode shortens that to 300 ms. It is not the standard condition, so it keeps its own record and shows no percentile. If your system asks for reduced motion, either mode slows to 700 ms per square.

What exactly is my score?

The last level you finished. If you complete a six-step sequence and then miss one square in the seven-step one, your score is 6. There is no partial credit.