The overarching goal of this thesis is to argue that conditionals, more specifically, conditional antecedents, are definite descriptions for the referential analysis of conditionals, cf. Schein 2001, Schlenker 2004, Bhatt and Pancheva 2006, among many others). So, in the same way that referential expressions like the mouse pick out an entity that is a mouse under the classical Fregean view, conditional antecedents like if Jerry was offered some cheese pick out (a set of) hypothetical scenarios where Jerry was offered some cheese. I show that conditionals pattern with definite descriptions in three respects that have not been explored in detail: number (singularity and plurality), semantic anaphoricity and discourse anaphoricity. First, definite expressions in many languages are marked for number. I argue that some conditional antecedents can be number-marked, too, as diagnosed by the interaction between Japanese conditional connective moshi ‘in case’ and various types of quantificational operators. In addition, definite expressions are known to have an anaphoric use, under which they pick out the individual that has been introduced by a preceding expression (e.g. in the discourse I met a man yesterday. The man told me a story, the definite the man picks out the same individual that was introduced by a man). I recognize a form of conditionals whose antecedents are anaphoric, as diagnosed by the distribution of wh-questions about elements inside conditional antecedents. Lastly, many expressions are known to display sensitivity to certain components of discourse (what I call “discourse anaphoricity”), such as the shared beliefs of conversational participants and the issues that are currently under discussion. I investigate the discourse anaphoricity of Japanese nara, which can be used to introduce both topics and conditional antecedents, and show that it displays similar discourse-anaphoric properties under its topic and conditional use.