The "reliability" levels reveal more about the inherently biased beliefs of the most active users (i.e., editors) than about the sources themselves. It turns out that "facts" are surprisingly slippery.