If questions are at the heart of creating new knowledge structures, then it follows that they are also at the heart of understanding in general. We have been working on programs that understand natural language text for many years now in our laboratory at Yale and at the Stanford AI lab before that. In general, the paradigm used in that work was to attempt to match an input string to some pre-existing framework. In the early days that framework was Conceptual Dependency (Schank, 1972,1973). Later we began to develop static knowledge structures and mapped sentences into those structures.