This paper presents a model for hypothesizing and tracking the changing task-level goals of a speaker during the course of an information-seeking dialogue. It allows a complex set of domain-dependent plans, forming a hierarchial structure of component goals and actions. Our model builds the user’s plan as the dialogue progresses, maintains both a local and a global plan context, and differentiates between past goals and goals currently pursued by the user. This research is part of a project to develop a robust natural language interface. If an utter-ance cannot be interpreted normally or a response cannot be generated due to pragmatic overshoot, the strong expectations about the utterance pro- vided by our context model can be used as an aid in processing the input and producing useful responses.