Working memory encoding of events and their participants: a neural network model with applications in sensorimotor processing and sentence generation

Abstract

In this paper we present a model of how events and their participants are represented in semantic working memory (WM). The model's central assumption is that events are experienced through sequentially structured sensorimotor (SM) routines - as are the individuals that participate in them. In the light of this assumption, we propose that events and individuals are stored in semantic WM as *prepared SM routines*. This proposal allows a new mechanism for binding representations of individuals to semantic roles such as AGENT and PATIENT. It also enables a novel account of how expectations about forthcoming events can influence SM processing in real time as events are perceived. Finally, it supports an account of the interface between semantic WM representations and language.


Back to Table of Contents