The two “relevance” criteria set out by Anderson and Belnap are discussed. It is argued that the motivation backing the variable sharing property is far weaker than it is commonly made out to be, and that the use-criterion does not distinguish between relevant logics such as E and R and “irrelevant” logics such as S4, intuitionistic and classical logic. In short, then, the paper argues that Anderson and Belnap’s two criteria of relevance are both motivationally unsound, and do not accomplish what they are commonly taken to accomplish.