For about a hundred years the Oxford English Dictionary said a siphon works because the air pushes the liquid up the short leg. The entry was written in 1911, and in 2010 a physicist in Queensland noticed it, checked every dictionary he could find, and found none that put gravity first. So he took a siphon into a low-pressure chamber and climbed it, on paper, to forty thousand feet. The flow barely changed. At forty-one thousand the column broke in two at the top, and when he brought the pressure back down it rejoined and carried on as if nothing had happened. The air turns out to matter mostly for keeping the water from boiling apart at the bend; the pull is the long leg falling. What I like is that the old definition was not stupid. It was the explanation that sounds right if you have only ever watched a siphon at sea level, which is everyone, which is why it sat in the book unbothered for a century. You only find out what was doing the work by taking away the thing that was always there.