Unnatural language discourse: an empirical study of multimodal proof styles

Computer-based logic proofs are a form ofùnnatural' language discourse, but the structure and process of proof generation can be observed in considerable detail, and analysis is leading to a number of general insights. We have been studying how students respond to multimodallogic teaching. Performance measures have already indicated that students' pre-existing cognitive styles have a signiicant impact on teaching outcome. Furthermore, a large corpus of proofs has been gathered via automatic logging of proof development. This paper applies a series of techniques, including corpus statistical methods, to the proof logs. The results indicate that students' cognitive styles innuence the structure of their logical discourse, via their diiering methods of handling abstract information, and transferring information between modalities. As well as uncovering diierent thinking styles in this artiicial domain, the observations raise the issue of the importance of individual diierences in natural language discourses.