My first question upon hearing this was, “How did they figure that out?” because I couldn’t really see an examiner sitting down with a bee to ask them to list a few numbers. It turns out (according to The Australian, found via the story on Yahoo! News) that Professor Mandyam Srinivasan placed landmarks in tunnels. If the bees were trained to go to the first landmark (where “trained” means that they discovered that nectar was always placed there), then they went to the first landmark to look for more nectar. If bees were trained to go to the second landmark, then that’s where they went to look for the treats. Same thing when the bees were trained to go to the third or fourth landmark.
But four was it: they couldn’t be trained to go to a fifth landmark. If there were more than four landmarks, they looked at all of them equally in their quest for sweet sweet nectar. (Note: it’s not clear to me if they were able to go to the first landmark if there were more than five, or if having more than five landmarks just completely confused them.)
Incidentally, a different news story in The Telegraph earlier this year announced that North American mosquito fish can also count to four. The fish could also distinguish between groups of very different sizes (16 versus 8, but not 16 versus 12). No word on how bees would do at the dots test.