This thesis reports on research undertaken in an exploration of the use of formal semantics for student modelling in intelligent tutoring systems. The domain chosen was that of tutoring programming languages and within that domain Prolog was selected to be the target language for this exploration. The problem considered is one of how to analyse students' errors at a level which allows diagnosis to be more flexible and meaningful than is possible with the 'mal-rules' and 'bugcatalogue' approach of existing systems. The ideas put forward by Robin Milner [1980] in his Calculus of Communicating Systems (CCS) form the basis of the formalism which is proposed as a solution to this problem. Based on the findings of an empirical investigation, novices' misconceptions of control flow in Prolog was defined as a suitable area in which to explore the application of this solution. A selection of Prolog programs used in that investigation was formally described in terms of CCS. These formal descriptions were used by a production rule system to generate a number of the incomplete or faulty models of Prolog execution which were identified in the first empirical study. In a second empirical study, a machine-analysis tool, designed to be part of a diagnostic tutoring module, used these models to diagnose students' misconceptions of Prolog control flow. This initial application of CCS to student modelling showed that the models of Prolog execution generated by the system could be used successfully to detect students' misunderstandings. Results from the research reported here indicate that the use of formal semantics to model programming languages has a useful contribution to make to the task of student modelling.