This paper presents a framework for the design and analysis of introductory statistics courses. This framework logically precedes the usual process of putting together the syllabus for an introductory statistics course. Four approaches, or paradigms, of statistics teaching are put forward, together with tools for deciding which blend of approaches is most useful in any particular case. These approaches do not correspond to the two traditional schools of thought in statistics education -- probability-driven or data-driven -- but rather constitute a new approach.