Academic journal article Memory & Cognition

Does a Pear Growl? Interference from Semantic Properties of Orthographic Neighbors

Academic journal article Memory & Cognition

Does a Pear Growl? Interference from Semantic Properties of Orthographic Neighbors

Article excerpt

In this study, we investigated whether semantic properties of a word's orthographic neighbors are activated during visual word recognition. In two experiments, words were presented with a property that was not true for the word itself. We manipulated whether the property was true for an orthographic neighbor of the word. Our results showed that rejection of the property was slower and less accurate when the property was true for a neighbor than when the property was not true for a neighbor. These findings indicate that semantic information is activated before orthographic processing is finished. The present results are problematic for the links model (Forster, 2006; Forster & Hector, 2002) that was recently proposed in order to bring form-first models of visual word recognition into line with previously reported findings (Forster & Hector, 2002; Pecher, Zeelenberg, & Wagenmakers, 2005; Rodd, 2004).

Researchers have been interested for quite some time now in how reading a word proceeds from the initial processing of its visual features to the activation of its meaning. Various models have been proposed in order to account for semantic processing of visually presented words. Strictly form-first models propose that a word's semantic properties are activated only after orthographic processing has completed and a unique word has been identified and selected for further processing (Forster, 2006; Forster & Hector, 2002). In these models, presentation of a target word activates not only the orthographic representation of the target word itself, but also that of orthographic neighbors.

Processing at the orthographic level continues until a unique word has been identified. Only then is information from this level used as input for processing at the semantic level. Cascaded models (see, e.g., Becker, Moscovitch, Behrmann, & Joordens, 1997; Masson, 1995; McClelland & Rumelhart, 1981), on the other hand, assume that information cascades from one level to the next as soon as processing has started. In these models, too, presentation of a target word activates the orthographic representation of the target word itself, as well as that of orthographic neighbors. However, in contrast to form-first models, as soon as a word's orthographic representation starts being activated, information from the orthographic level is used as input for other levels. Thus, semantic information gets activated before the processing of orthography has been completed.1 The distinction between form-first and cascaded activation is fundamental to models of word recognition. For example, cascaded activation is a necessary condition for certain effects of feedback processing on visual word recognition (Hino, Lupker, & Pexman, 2002; Pecher, 2001; Pexman & Lupker, 1999; Pexman, Lupker, & Hino, 2002; Stone, Vanhoy, & Van Orden, 1997; Van Orden & Goldinger, 1994; Ziegler, Montant, & Jacobs, 1997).

Evidence for cascaded processing is provided by studies that show that semantic properties of a target's neighbors affect responding to the target word itself (Bourassa & Besner, 1998; Duñabeitia, Carreiras, & Perea, 2008; Forster & Hector, 2002; Pecher, Zeelenberg, & Wagenmakers, 2005). Forster and Hector presented words and nonwords in an animal decision task. They observed that categorization was slower for nonwords (e.g., turple) that had an animal neighbor (e.g., turtle) than for nonwords (e.g., cishop) that did not have an animal neighbor. These results are consistent with the view that semantic properties of the neighbor are activated before the target is recognized as being a nonword (but see below for an alternative interpretation). Had orthographic processing been completed before the activation of the semantic information, the semantic properties of the neighbor would not have interfered with the decision.

An important characteristic of the Forster and Hector (2002) study is that the critical manipulation was done with nonword stimuli. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.