| Two Event-Related Potential (ERP) studies and one behavioral study are presented in service of a single goal: to evaluate current prevalent computational models of visual word recognition. Specifically, the functionally homogenous cooperative division of labor model (Triangle) and the full family of dual-route models (Dual-Route Cascaded and Connectionist Dual Process Plus) are contrasted. The results of the three experiments unanimously suggest that visual word recognition is accomplished via functionally homogenous computations that are not dependent on factors such as orthographic regularity---results which contradict core tenants of dual-process models and are instead more consistent with functionally homogenous models, such as the Triangle model. In addition to this support for Triangle-type models, two other conclusions are drawn. First, that orthographic neighborhood size (i.e., Coltheart's N) is a considerably more important determiner of ERP morphology than has previously been thought, and second that the N400 ERP component can be elicited by a considerably wider range of orthographic inputs than previous studies have suggested. |
