Abstract
According to contextualist accounts, the truth value of a given knowledge ascription may vary with features of the ascriber's context. As a result, the following may be true: "X doesn't know that P but Y says something true in asserting 'X knows that P'". The contextualist must defend his theory in the light of this unpleasant but inevitable consequence. The best way of doing this is to construe the context sensitivity of knowledge ascriptions not as deriving from an alleged indexicality of the word "know" nor from its vagueness or ambiguity, but rather from a distinct semantic feature of the word "know", namely its unspecificity.