Measured in squares remembered ยท higher is better
Chimp Test
Click the numbers in order from 1. After your first click they turn blank. One more square every round.
Remember where the numbers are
Numbered squares appear on a grid. Click them in order from 1 up. From the second level on, the numbers turn blank after your first click, so remember where each one was. One more square every time you clear a board. Three strikes ends the run.
Before you start
The median on mouse is 10 squares. Run the test and we will place you against it.
Compared against
There is no clock in this test, but squares are smaller on a phone and a tap can land on the wrong one. Ranked separately by device.
The other device classes
Trackpad
Touchscreen
Curve is a fitted baseline plus 0 first attempts recorded here. The baseline is a curve, not a sample: no part of it is a real person. Baseline fitted from Human Benchmark's chimp test score chart (n=27,233) (source). Human Benchmark shows this chart without a date or sample size. The counts come from the data behind the chart (27,233 scores, median 10, mean 10.3); we fitted a normal curve to them, and the split by input device is ours. The baseline is dropped once a device class holds 5,000 real first attempts. See the full distribution.
Understanding your score
| Band | squares | What it means |
|---|---|---|
| Exceptional | 16 and above | About 4 in 100 scores. Almost always a deliberate path or grouping strategy. |
| Very good | 13 to 15 | About 9 in 100. 13 is the 90th percentile of the Human Benchmark scores. |
| Good | 11 to 12 | Above the median of 10. About 28 in 100 scores. |
| Average | 9 to 10 | Where about 4 in 10 scores land. The median is 10. |
| Below average | 7 to 8 | Common on a first run while the rules are still new. |
| Low | 6 and below | Usually a rushed first click. Look at the whole board before you start. |
How the chimp test works
Numbered squares appear at random positions on a grid of eight columns and five rows. Click them in ascending order, starting at 1. On the first level the numbers stay visible the whole time, so it only teaches the rules. From the second level on, every number turns into a blank white square the moment you click square 1, and the rest has to come from memory.
The run starts with four squares. Clear a board and the next one has one more. Click a square out of order and you take a strike, and a fresh board of the same size is dealt. Three strikes end the run. Your score is the number of squares on the board where the run ended, so a run that clears the boards with four, five, and six squares and then strikes out three times on seven scores 7.
Between boards you see the current count and your strikes, and nothing starts until you press Continue. There is no clock anywhere in this test. You can look at the board for as long as you like before you click 1, and you can take as long as you like over every click after it.
Where the name comes from
The test is named after, and modelled on, a study by Sana Inoue and Tetsuro Matsuzawa at the Primate Research Institute of Kyoto University, published in 2007 as "Working memory of numerals in chimpanzees" in Current Biology (volume 17, issue 23, pages R1004 to R1005). Six chimpanzees, three mothers and their offspring, learned to touch the Arabic numerals 1 to 9 in order on a touch screen. The young ones began this training in 2004, at four years old.
In the masking task, all the other numerals were replaced by white squares as soon as the first numeral was touched, and the chimpanzee had to touch the squares in the order of the numbers that had been there. That is the mechanic on this page. The best performer was Ayumu, the son of Ai, a chimpanzee who had earlier learned to label sets of objects with Arabic numerals.
The comparison that made the study famous was a second version, the limited-hold task. After the subject touched a start circle, five numerals appeared for only 650, 430, or 210 milliseconds and were then replaced by white squares automatically. Nine university students took the same task. Their accuracy fell as the display got shorter, and Ai's did too. Ayumu's stayed at about the same level at all three durations, and the authors report that he outperformed all of the human subjects in both speed and accuracy.
The paper also states the conditions. Each chimpanzee received ten sessions of 50 trials; each of the nine humans received a single session. The authors compare the result with eidetic imagery, a capacity to retain a detailed image that is reported more often in children than in adults.
This version is not the lab task
The browser chimp test keeps the masking idea and drops the time limit. Nothing here disappears on a timer: the numbers stay up until you click 1, however long that takes. The shortest display in the study, 210 milliseconds, was chosen because it is too short to move your eyes across the screen. Here you can scan the board as many times as you want.
That changes what is being measured. The lab result was about how much you can take in at a glance. This test measures how many positions you can hold in order once you have had time to study them, which is ordinary visuospatial working memory plus whatever strategy you bring to it. A good score here says nothing about how you would do against Ayumu at 210 milliseconds, and the study says nothing about how a chimpanzee would do on this untimed version.
What actually improves the score
Trace a path. Before clicking 1, follow the numbers with your eyes from 1 to the last one and treat the route as a single shape. A zigzag across the grid is easier to hold than ten separate coordinates.
Group in threes or fours. Hold 1 to 4 as one cluster, 5 to 8 as the next, and so on. Most people who pass 13 report some kind of grouping, and the grouping does not need to be tidy.
Spend the time before the first click, not after it. Once the numbers are blank, staring at the board does not bring them back. All of the useful looking happens while they are still visible.
Do not let the click order run ahead of your memory. Each strike deals a completely new board, so a careless click costs you everything you had memorised for that board, not just one square.
What the score does and does not tell you
A score of 9 to 10 is ordinary. On Human Benchmark's own chimp test, the data behind its score chart, read on 22 November 2026, holds 27,233 scores with a median of 10 and a mean of 10.3. The middle half of those results runs from 9 to 11, and only about 1 in 100 reaches 20.
Like the other span tests on this site, the chimp test is narrow. It correlates with other working memory measures and says little about remembering names, appointments, or where you put your keys. Sleep, stress, and distraction can move a result by several squares from one day to the next, so one run is a reading of this attempt, not a verdict on your memory.
Questions
- What is the average chimp test score?
- About 10 squares. The data behind the score chart on Human Benchmark's chimp test page, accessed 22 November 2026, holds 27,233 scores: median 10, mean 10.3, 10th percentile 8, 25th percentile 9, 75th percentile 11, and 90th percentile 13. Those figures are read from the chart data, not from a published study.
- What is a good chimp test score?
- 13 is the 90th percentile of the Human Benchmark scores, and about 1 in 8 of them is 13 or higher. 16 or more is about 4 in 100. 20 or more is roughly 1 in 100 and almost always comes from a deliberate strategy such as tracing a path through the numbers.
- Why do my numbers disappear?
- That is the test. From the second level on, every number turns into a blank square as soon as you click square 1, so the rest of the order has to come from memory. On the first level the numbers stay visible so you can learn the rules without a penalty.
- How is this different from Human Benchmark's chimp test?
- The rules are the same: an 8 by 5 grid, four squares to start, one more after each cleared board, three strikes, and the score is the number of squares on the board where you stopped. So the scores compare directly. The difference is the ranking. Your percentile here is split by input device, and once a device class holds enough real first attempts on this site, it is ranked against those instead of the fitted baseline.
- Is the chimp test a good test of memory?
- It is a reasonable test of one kind of memory: holding a handful of positions in order for a few seconds. It is not a clinical instrument, it is easy to improve with practice, and the gains come mostly from strategy rather than from a larger memory. If your memory has changed noticeably in daily life, tell a clinician; a browser score cannot answer that question.
- Did a chimpanzee really beat humans at this?
- At a different version of it. In Inoue and Matsuzawa's 2007 study, the young chimpanzee Ayumu kept his accuracy when five numerals were shown for as little as 210 milliseconds, while nine university students lost accuracy as the display got shorter. The chimpanzees had ten sessions on that task and each student had one. This page has no time limit, so it does not repeat that comparison.
- What happens after three strikes?
- The run ends and your score is the number of squares on the board you were on. A wrong click never shrinks the board: you get a fresh layout of the same size, so three strikes on seven squares scores 7.
- Can I play with the keyboard?
- Yes. Every square is a button. Tab moves between squares in grid order, and Enter or Space clicks one. Once the numbers are hidden, a screen reader announces a square by its row and column only, never by its number.
Where does the curve behind this page come from, and when does it get thrown away? The methodology page names the source of every reference distribution on the site, the sample size behind it, and the point at which real results replace it.