I look at some of the classic problems in characterizing the semantics of the disjoint reference effect that Principle B governs. I look at popular accounts of this that make use of Tanya Reinhart's idea that a key ingredient involves comparing different ways of representing a meaning. I argue that this idea works but only if the different ways of representing the meaning are all syntactically well-formed. I suggest that what is relevant is not, as has sometimes been suggested, comparing ways of referring to the entities described in some sentence, but is instead comparing ways of individuating the situations that those entities are parts of.