This dissertation is about the semantic treatment of indexicals, that is, linguistic elements that refer to properties of the context in which an utterance is produced, such as I and you. Since Kaplan (1977), philosophical theorizing has it that indexicals differ from other noun phrases because they are directly referential: in Fregean terms, they do not obtain their reference by virtue of their sense (Sinn), but by referring directly to the objects they stand for (their denotations, Bedeutungen). As a consequence, indexicals are therefore set apart from other nominal or adverbial elements of the same type, which all have a sense through which reference is achieved. This conclusion, however, is at odds with some other assumptions linguists entertain about natural languages. For instance, it is generally admitted that first and second person indexicals are personal pronouns and, as such, are referential elements carrying abstract features that pattern rather uniformly at the morphosyntactic level across the world's languages (Adger and Harbour, 2008; Bobaljik, 2008); on the assumption that interpretation is mediated by the morphosyntax, should we not expect these elements to behave uniformly at the interpretive level as well? Through the study of various phenomena involving indexicals in a rich body of natural language data, this dissertation aims at vindicating the hypothesis that some semantic uniformity is to be found between indexical and non-indexical personal pronouns. More precisely, looking at attitude reports cross-linguistically, we investigate four phenomena that all encourage a reassessment of some of Kaplan's claims about indexicals: logophoric pronouns, shifted indexicals, sign language indexicals under role shift, indexicals in quotation, and indexicals under ellipsis.