Proceedings of the Annual Meeting of the Cognitive Science Society

Propositional accounts assume sentences are encoded in terms of a set of arguments bound to role-fillers in a predicate,but they never specify how the role representations form in the first place. Dennis (2005) shows an alternative way tocapture role-information based on simple associations derived directly from experience in the Syntagmatic-Paradigmatic(SP) model. We argue that the evidence for the propositional view is not well-founded and explore the possibility for apure associative encoding of proposition-like information. We differentially manipulate overlap in target and distractorsentences, embedded in narratives, and directly place the propositional account against the SP view. Our first experimentprovides some evidence for an SP account, however the second experiment supports the propositional view. Our finalexperiment provides results that are difficult to explain with either account. Overall, our results support the propositionalview and show mixed evidence for the SP account.