The chief aim of the article is to give an analysis of Russellś semantical theory of ordinary proper names with respect to the problem of reference. The problem can be stated in the following way: proper names are employed as devices for identifying objects - how is it possible that they can serve for this purpose? It is argued that Russellś semantics is unable to solve the problem and, moreover, it is impossible to state the problem within its conceptual framework. Russellś identification of the referent of a name with nameś meaning, together with his logico-semantical distinction between complete and incomplete symbols, and his epistemology of sense-data implies that grammatically proper names are, in fact, disguised descriptions. It is argued that in order for the theory to work it must rely on an infallible or ideal speaker.