Perea Neighborhood Effects in Visual Word Recognition and Reading 6

76 Introduction The examination of the nature of the underlying mechanisms that associate a printed word with its correct lexical unit (i.e., the process of lexical access) is one of the most basic issues in the research on reading. There are several reasons for this relevance. First, lexical access is a central component of sentence reading (Besner & Humphreys, 1991). Second, many reading disorders may originate from a deficient process of word identification (e.g., see Castles & Coltheart, 1993). There is some agreement that when we identify a word in an alphabetic language, there is an early stage at which a number of similarly spelled lexical units to the printed stimulus (i.e., neighbors) are partially activated (or accessible). That is, during the process of visual word recognition there is a collection of lexical candidates that are similar (in some sense) to a given word and these candidates influence the ease with which the stimulus word is encoded or perceived. During the course of word processing, these lexical candidates are progressively deactivated until only one lexical unit remains active (i.e., the perceived word) (e.g., search model: Murray & Forster, 2004; Bayesian reader model: Norris, 2006; interactive activation model: McClelland & Rumelhart, 1981, and its successors: multiple read-out model: Grainger & Jacobs, 1996; dual route cascaded model: Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001; spatial coding model: Davis, 2010). It is important to note here that there are other models that have a completely different metaphor for word identification, as in parallel distributed processing (PDP) models (Seidenberg & McClelland, 1989; see also Woollams, this volume). The present chapter first examines the different metrics that have been proposed to define the set of a word’s neighbors. It then examines whether a word’s neighbors help or hinder during the process Abstract