The chimpanzee memory test

The first two rows are as a later paper read them from the original study's chart; the original printed the results as a figure, not as numbers^[2]. The last two rows come from the study that trained the students^[3].

The study came from the Primate Research Institute of Kyoto University. Six chimpanzees, three mothers and their young, first learned the order of the numbers 1 to 9 on a touch screen. Training began in 2,004, when the young were four years old^[1].

In the main memory task, a set of numbers appeared in random places, and as soon as the chimpanzee touched the first, all the others turned into white squares. To succeed it had to remember which number had been where and touch the squares in order. Guessing almost never works: with five numbers there is one right order in 120^[1]. All five chimpanzees new to the task learned it, and the young ones did better than their mothers.

To compare species fairly, the researchers then showed five numbers for a fixed time only: 650 ms, 430 ms or 210 ms. The shortest is about as long as one quick look, too short to move the eyes around the screen. The students got worse as the time got shorter. Ayumu's accuracy hardly changed, and he beat every student in both speed and accuracy^[1].

The comparison had one large imbalance. Each chimpanzee had done 10 sessions of 50 trials on the timed task alone, after months of training on the untimed one; each student did a single session^[1].

Two researchers tested that directly. In a paper published in 2009, they practised the timed task themselves, aged 63 and 33, in session after session; one completed 84. Both reached Ayumu's level of accuracy at 210 ms^[2].

A second team went further and copied the chimpanzees' training. The chimpanzees had practised with numbers they could look at for as long as they liked, which in practice was about 650 ms; so two students practised at that duration, 50 trials at a time, until they had done 300 blocks. Before training they managed 38% and 46% at 210 ms, much like the original students. After it, 94% and 96%, better than Ayumu. The authors concluded there is no evidence that chimpanzees have a better or different kind of spatial memory^[3].

The Kyoto researchers suggested a reason why the young chimpanzees did so well. Their results, they wrote, may be reminiscent of eidetic imagery: the ability to keep an exact picture of a scene for a short while, which has been described in some children and is said to fade with age. The young chimpanzees in their study did better than the adults, which fits that idea^[1].

The team that trained the two students tested the idea in a simple way. Besides the usual white squares, they covered the numbers with a busy pattern of lines that would wipe out any picture left on the eye. Their students' accuracy fell to 77% and 81.5%. The authors read this as a sign that the task relies largely on ordinary visual memory, the brief trace an image leaves after it disappears, and noted that whether eidetic imagery exists at all is itself disputed^[3].

Nobody has run the same pattern test on Ayumu in the papers we read, so the question of exactly how he did it is still open. What the studies agree on is the simpler point: given enough practice, people can do the task very well too.

Both findings are true at once. Ayumu really was remarkable: he was the best of six chimpanzees, and he took in a screen of numbers at a glance faster than untrained adults could. But the task rewards a skill that improves hugely with practice, and once people practised as much, they did as well or better.

The study is still a good reminder that memory is not one thing. Holding a picture of a screen for a moment is different from remembering a list of numbers read out to you, and different again from remembering a story. People who are good at one are not always good at the others.

It also says something about tests in general. A test measures the skill and the practice together. Two people with the same memory can score very differently if one has played the game a hundred times. That is worth keeping in mind whenever a score is compared with someone else's.

Our chimp test is built on the same idea as the masking task, without a time limit. Numbered squares appear in random places, starting with 4. When you touch the 1, the rest are hidden, and you touch them in order from memory. Each round you complete adds a number; mistakes cost a strike.

• The score is the most numbers you remembered in one round.
• Because the numbers stay on screen until your first touch, you can take your time to look, as the chimpanzees could in training.
• Your result is compared with your own earlier ones: no study gives an average for this exact game.
• Expect to improve with practice, as the students in the studies did.

How many items people can usually hold at once, and why practice changes it, is explained in memory span.