Research Talk: Capturing Single Item Word Reading ERP Data in a Computational Model

Abstract

Parallel Distributed Processing (PDP) models have been widely used for modeling cognitive tasks where accuracy or reaction time were the dependent performance measures. However, only few PDP models have attempted to model more brain-related data like event related potentials (ERPs). In this presentation, we propose a PDP model, which can successfully replicate various known ERP effects. Specifically, we introduce a PDP-equivalent of the N400 ERP measure and apply it to a simple PDP model of early bilingual word acquisition as bilingual word acquisition tasks provide several well- established N400 effects that can be used for model validation. We then analyze the dynamics of the network to show why and how the network can capture each of the targeted N400 effects. We close the presentation by discussing the critical properties of a minimal PDP architecture capable of replicating the ERP data.