The aim of this paper is to provide an analysis of the function of imperatives in dialogue. In particular, the focus is on the use of logically complex imperatives (e.g., `Say hello to John, if you meet him') and temporal reference in imperatives (e.g., `Open the bottle of champagne at midnight'). Two specific problems involving logically complex and temporally referring imperatives are introduced. These problems are addressed within a framework for communicating agents. It is argued that such a framework needs to allow for partial models of communicating agents and their environment and for the intentionality of agent states.