The bee paper referenced in the original article (Dacke & Srinivasan 2008) is about real bees, and does a pretty good job of showing that the bees must somehow

The bee paper referenced in the original article (Dacke & Srinivasan 2008) is about real bees, and does a pretty good job of showing that the bees must somehow be keeping numerical track of how many instances of a particular landmark they've passed, at least up to a total of four.