Chimp test with the exposure the chimpanzees actually got

At the timed version, one particular chimpanzee beat a group of university students, and that is a narrower claim than the internet made of it. Ayumu, then about five years old, was tested at the Primate Research Institute at Kyoto University and held roughly 80% accuracy whether the numerals were covered after 650 ms or after 210 ms, while the student group fell from around 80% to about 40% over the same range. At the self-paced version — the one you are most likely to be playing — there was no comparable gap, because a human who can look as long as they like will look as long as they need.

Because it removes the part you were actually good at. Self-paced play lets you rehearse the layout in chunks, check yourself, and only then commit by touching the 1 — the page shows your median study time so you can see how much of your score that was buying. At 210 ms nothing can be rehearsed, since the whole display is gone in about a third of the time it takes to move your eyes twice, so the round is decided by one glance. That is a measure of how much you can take in at once rather than how well you can hold it.

Nine self-paced is a solid result and most people stall between seven and eleven. The reason nine keeps coming up is that the original task used the numerals 1 to 9, so the ceiling was built into the apparatus rather than into the people. In the 210 ms mode expect the figure to fall by three or four numerals, and expect the fall to be steepest at exactly the count where you had been comfortable — the point at which one look stops being enough is quite sharp.

No, there is no clock running once the blocks are down. The page times how long you take from the mask to the last numeral and reports the median, but nothing in the scoring uses it; the only limits on the round are the order and the three strikes. Speed matters in one indirect way — a layout is easier to hold for four seconds than for forty — but hurrying your clicks buys accuracy problems rather than points.

Because letting you guess again would turn twenty cells into a coin-flip exercise. With eight numerals hidden among 42 cells, a policy of guess-and-retry clears most rounds eventually and the count you reach would then describe your patience. Ending on the first wrong numeral keeps a cleared round meaning what it appears to mean: that the sequence was reproduced from memory in one pass. The three strikes are what give you room for a genuine slip.

No — every round draws a fresh set of cells at random and assigns the numerals to them in a fresh order. This is deliberate rather than incidental: a fixed or repeating layout is learnable, and a few rounds in you would be recalling a position from the previous attempt instead of encoding this one. It also means two rounds at the same count can differ in difficulty, since numerals bunched in one corner are considerably easier than numerals spread to the edges.

Because a browser can only change the screen on a frame boundary. The blocks are drawn on the first animation frame at or after the target, and frames arrive every 16.7 ms on a 60 Hz display, so a 210 ms request is served as anything from 210 to about 227 ms; at 144 Hz the overshoot shrinks to under 7 ms. Reporting the hold you were actually given rather than the one you asked for is the honest version, and it is also why a result from a 60 Hz laptop is very slightly easier than the same setting on a fast monitor.