An exciting subject of current investigation in AI and Cognitive Science today is the search for representations of knowledge that can shed light on the relationships between diverse human capabilities, for example the abilities to read and to draw. The SAT computer program is a natural language understanding program. 1 The purpose of the program is to take English narrative accounts of traffic accidents as input; to analyze them, build a concrete model, and actually show an animation of the event. The SAT program is therefore at its core concerned with representation of actions involving both space and time. The program has already had some initial success with a small set of narratives. This paper focuses on describing the set of internal representations that were successfully used in SAT to take us from English sentences to a graphical animation.