Automated Student Modelling explicates reasoning strategies, the representation of procedural skills, and underlying misconceptions as manifested in errors. A diagnostic model, based on a "procedural network" as opposed to a "semantic network," is presented which provides a technique both for modelling the underlying cognitive processes of a procedural skill and for finding a way to account for manifested errors in the performance of that skill. A technique is described for analyzing the problem solving trace or protocol of a student and then automatically synthesizing a model of the problem solving strategies and motivations which were used to arrive at the solution. The underlying theory captures the reasoning powers of a master tutor and is seen as useful for guiding a computer-based laboratory tutor and for measuring how problem solving performance is evolving. It differs from classical computer assisted instruction by focusing on techniques for teaching procedural knowledge and reasoning strategies which are best learned through hands-on tasks with the guidance of an automated intelligent tutor. (Author/DAG)