What do External Representations Tell about Mental Models? An Exploratory Study in Deductive Reasoning

We present an exploratory study in deductive reasoning based on the experimental elicitation of external representations of the premises as a tool to investigate how individuals build models of the premises and how they use them in reasoning. The goal of our research is not to check whether external, explicit models are useful as an heuristic support in deductive tasks, but to devise a methodology which allows to make as explicit as possible the mental models that people build when executing deductive tasks. We show that while the number and the completeness of models constructed is not significantly associated to deductive performance, the quality of models constructed does matter. Further analysis of external representations lends support to some predictions of the theory of mental models regard which models are more likely to be constructed. In addition, the qualitative analysis of data gave us some interesting information on typical errors patterns, which we clustered in three main categories: model editing, model integration, and modalization.