In verbal communication we constantly use singular terms with the intention to refer to certain objects we have in mind. The central question that I will be concerned with in my thesis concerns the conditions under which such referring acts are successful. That is, what must be the case such that a hearer can be credited with understanding the referential use of a singular term by some speaker? Like other writers I locate the interest in this question in the following tension that arises with regard to it: on the one hand it seems that the object referred to in a referring act should be constitutive for communicative success, that is agents will succeed in this form of communication if and only if their underlying thoughts refer to the same object in the world. But on the other hand it seems that communicative success in referring acts cannot be object-dependent in such a way, as examples similar to Frege's Hesperus-Phosporus one or certain 'empty' cases like children's uses of the name 'Santa Claus' show. The aim of my thesis is to come up with a satisfying account of success in referring acts which resolves this tension. In this regard various traditional and recent accounts will be discussed, in particular the prominant Fregean one which requires identity in the entertained modes of presentation for communicative success, and Evans' (1982) hybrid account that combines an external success condition with a Fregean one. Yet it will be argued that all those accounts fail on descriptive grounds. An alternative account of communicative success will be proposed which in some aspects closely resembles Evans' one but which also departs from it in crucial ways. First, a different Fregean success condition will be appealed to, and secondly also the distinction between cases of referential communication which are object-dependent and those which are not will be drawn along different lines. Perception-based cases where one intends to refer to an object currently perceived are considered to be object-dependent, whereas other cases, in particular the communication-based ones where one intends to refer to an object one has heard of from others, are not treated in such a way. The resulting account will allow us to overcome the problems faced by previous accounts.