Mental models and local semantics: the problem of information integration

In this paper, we sketch a framework complementing the theory of mental models with the notion of context. A context is defined as a collection of local mental models, namely mental models which represent only part of a situation. We argue that most reasoning tasks essentially involve a stage at which the information represented in a set of initial “input” contexts is integrated in a more comprehensive “target” context, where a solution can be searched for. Our claim is that this more structured account of mental representations can provide insights in the constructive process through which mental models are generated and assembled. The preliminary experimental evidence reported in this paper shows that the organization of contexts matters in explaining how individuals perform disjunctive reasoning tasks, and that the difficulty in information integration complement the number of mental models as an explanation for deductive failures.