A Dynamic Update Model of Sentence Processing

In recent years formal semantic theories based on the principle of compositionality have been criticized of not being able to explain how meaning is computed in the brain (Baggio and Hagoort 2011). For example, although neither John squeezed an orange nor John squeezed an apple contain a semantic anomaly, they are processed differently in the brain, because orange is more expected as the direct object of squeeze than apple. Alternative approaches, like that of Van Lambalgen and Hamm (2005) abandon compositionality by allowing lexical meanings to be both enriched and reduced during semantic processing. In this article we try to bridge the gap between neuroscience and formal semantics. We develop a compositional dynamic update model of sentence processing that is based on strategies from Dynamic Epistemic Logic and Epistemic Temporal Logic. For example, the meaning of a verb is a set of evolutions, modeled by a branching tree that describes how events can evolve. The branches of the tree are ordered by a preorder that captures a comprehender’s conditional beliefs about the plausibility of the different evolutions. New information is modeled as a (doxastic) event model. The result of updating the current discourse model with this new information is defined as a form of priority update.