This paper shows that differential object marking (DOM) is not a unitary cross-linguistic phenomenon, and observes that its two subtypes, differential object flagging (DOF) and differential object indexing (DOI), probably have rather different explanations. The cross-linguistic patterns found with both subtypes of DOM can be described by the scales of referential prominence (animacy, specificity, and so on), but DOI is like differential subject indexing (DSI) in that it conforms to the prominent indexing universal, while differential subject flagging (DSF) shows the mirror-image pattern of DOF: Flagging is preferred when the subject is not prominent on one of the prominence scales. The paper provides examples from a wide variety of languages showing that indexing behaves rather differently from flagging. Thus, DOM is not a unified phenomenon, as was already noted by Siewierska (2004) and some earlier authors.