In this dissertation, I provide (i) a semantics for simple and complex singular demonstratives in English, (ii) a theory of how the reference of a referential use of a singular demonstrative is determined in the context of the use, and (iii) a logic of demonstratives that allows us to accurately represent our reasoning with demonstratives. The last project requires rejecting David Kaplan's original definitions of validity and logical truth for languages containing context-sensitive expressions; according to the alternative I propose, a sentence S can be logically true, or an argument valid, relative to one context but not another. This allows us to explain why we can use an argument containing demonstratives in one situation and seem to reason well, but use the same argument in a different situation and seem to reason poorly. Crucial to this view is to recognise the distinction between referential and non-referential uses of demonstratives, where referential uses are distinguished by a particular kind of referential intention with which the speaker uses the demonstrative. The semantic theory I propose accounts for both standard referential uses and deferred referential uses of demonstratives, while preserving the classic intuitions about direct reference for standard uses. I defend this semantic proposal against several objections and compare it to the most important alternatives currently available in the linguistic and philosophical literature.