This paper describes ongoing research in the development of formal techniques and methods for human agent interaction using a layered protocol based approach and a task communication language (TCL). This approach focuses upon the interaction between a dialogue manager and the behavioral aspects of an intelligent agent. This separation allows parallel development of each as well as interoperability through a flexible, dynamic and adaptive interface. We describe the proposed techniques for creating the semantics for a task communication language and modeling methodology and providing empirical evaluations. We show how this will allow for the merging of a variety of existing dialogue and interaction models into a single integrated system.