Ayumu and the white squares
At the end of 2007 the world's press was talking about a young chimpanzee. Ayumu, who lived at Kyoto University's Primate Research Institute, could point in order to where numbers had briefly appeared on a touch screen after they had turned into white squares, and he did it better than university students.
This experiment is a version of that test. Tap the circle in the middle and a few numbers between 1 and 9 appear in random places on the grid, then turn into white squares very quickly. Your job is to tap the squares from smallest to largest: first where the smallest number was, then the next. The time drops from 650 milliseconds to 210, and then the numbers multiply.
Inoue and Matsuzawa's experiment
Sana Inoue and Tetsuro Matsuzawa first taught six chimpanzees, three mother–offspring pairs, the order of the numerals 1 to 9. They then used two memory tests. In the 'masking' test, touching the first numeral covered all the others with white squares. In the 'limited-hold' test, touching a start circle made five numerals appear for a fixed time only, after which they were covered automatically: 650, 430 or 210 milliseconds.
650 ms was chosen to match the average time the chimpanzees spent before touching the first numeral in the masking test. 210 ms is close to the rate at which the eyes jump from one point to another, so there is no time to scan the screen. Ayumu's performance barely changed as the time got shorter. Nine university students who took the test in a single session got worse as it shortened; at 430 and 210 ms the difference between Ayumu and the students was statistically significant.
The numbers behind the comparison
The paper reported its results only as a graph. In their 2009 reply, Alan Silberberg and David Kearns read off these values: at 650 ms the students were as good as Ayumu, both completing about 79% of trials perfectly. At 210 ms Ayumu stayed at about 79% while the students' average fell to about 38%.
The third round of this game repeats exactly that condition: five numbers, 210 milliseconds. On the results screen you will see how many of the five trials in that round you completed perfectly, side by side with Ayumu's and the students' rates. Completing four of the five matches Ayumu's rate.
The critique: a practice gap
Silberberg and Kearns pointed out an important difference in method: Ayumu had played the task for months, whereas the students were seeing it for the first time. They played it themselves. Silberberg started recording without practice, close to the 2007 students' level, and reached Ayumu's level after about 2,500 trials. Kearns had played a lot while programming the task: he was at about 30% at 210 ms when he first tried it and was already at Ayumu's level by the time he started recording. In their view, the key to success in this task was not youth or species but practice.
In 2010 Peter Cook and Margaret Wilson went further. Before practice, two university students were 46% and 38% correct at 210 ms, much like the 2007 students. After weeks of practice at 650 ms, 15,000 trials each and similar to Ayumu's training, they reached 94% and 96% at 210 ms and significantly outperformed Ayumu. They also noted that Ayumu was the only one of the six chimpanzees to reach that level. The authors concluded that there is no evidence for a superior or qualitatively different spatial memory system in chimpanzees.
A trace in the eye, or extraordinary memory?
Inoue and Matsuzawa had linked Ayumu's performance to 'eidetic imagery', the ability to hold a detailed image of a scene in mind for a short time, reported to be more common in children. To test that interpretation, Cook and Wilson replaced the white squares with a pattern mask made of lines resembling parts of digits. Accuracy dropped from 94–96% to 77–81%.
Because a pattern mask disrupts the brief visual trace left in the eye (iconic memory or an afterimage), this result suggests the task relies largely on ordinary iconic memory. The most balanced reading today is that Ayumu really is an outstanding player, but it has not been shown that the ability is special to chimpanzees or comes from a system different from ours. The debate is a good example of why you should look at the conditions of a comparison before concluding that one species is 'smarter'.
What your score does and does not tell you
Each trial scores the number of squares you tapped correctly before your first mistake divided by the number of numbers: get three of five in order and then slip, and that trial scores 0.6. A round's score is 10 times the average, and the maximum total is 50. The key comparison on the results screen, though, is the share of perfect trials, because Inoue and Matsuzawa also counted only trials in which every number was ordered correctly.
Your rate is based on a handful of trials; a single wrong tap changes your 210 ms rate by twenty percentage points. Because of your screen's frame rate the numbers may stay up a few milliseconds longer than intended; we record the measured time. Replaying the test is practice too: as Silberberg and Kearns showed, you should expect to improve as you play. That is why your first game is the one kept in the scientific data.