Academic journal article Canadian Journal of Experimental Psychology

Neighbourhood Density, Word Frequency, and Spelling-Sound Regularity Effects in Naming: Similarities and Differences between Skilled Readers and the Dual Route Cascaded Computational Model

Academic journal article Canadian Journal of Experimental Psychology

Neighbourhood Density, Word Frequency, and Spelling-Sound Regularity Effects in Naming: Similarities and Differences between Skilled Readers and the Dual Route Cascaded Computational Model

Article excerpt

Abstract An experiment with skilled readers and a series of simulations with the Dual Route Cascaded model (Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001) investigated the joint effects of stimulus quality and Neighbourhood Density (N) in nonword naming. Neighbourhood Density and stimulus quality yielded additive effects on RT for skilled readers whereas the model produced an interaction between these factors. A further set of simulations show that DRC also produces an interaction between stimulus quality and (1) word frequency, (2) spelling-sound regularity, and, (3) nonword letter length. None of these three factors interact with stimulus quality in performance by skilled readers. It is suggested that DRC's assumption of cascaded processing throughout represents a central problem. A proposal as to how the model can be modified to accommodate these and other problematic data is discussed.

Theories of how humans read at the single item level are often concerned with accounting for, and making new predictions about, both the joint effects of factors on reaction time (RT) and the temporal processing sequence associated with these factors. The present paper begins by exploring the processing that underlies the effect of Neighbourhood Density (N) on nonword reading in both skilled readers and in Coltheart, Rastle, Perry, Langdon, and Ziegler's (2001) computational Dual Route Cascaded model (DRC). The locus at which the N effect arises is assessed by examining the joint effects of stimulus quality and N. The results of these investigations are consistent with the conclusions that (a) there is a single late locus for the N effect in non-word naming for skilled readers, and (b) the currently implemented version of the DRC model does not correctly simulate the N effect seen for skilled readers.

A further set of simulations shows that DRC produces a three-way interaction between stimulus quality, word frequency, and spelling-sound regularity. This is also problematic for the model given that stimulus quality has additive effects with word frequency and spelling-sound regularity in human naming performance.

We end with a discussion of the general claim that, among other difficulties, a major problem for the DRC model is the fact that processing is cascaded throughout. It is suggested that (a) thresholding the output from the letter level (rather than allowing it to cascade) and (b) restricting the role of feedback between levels will allow the model to successfully simulate data that are currently problematic.

Introduction

Computational models are sometimes preferred over "verbal models" because they are more complete and can be directly tested (e.g., Besner & Roberts, 2003; Coltheart et al., 2001; Jacobs & Grainger, 1994; Reynolds & Besner, 2002). When the model does not successfully simulate human performance, attempts can then be made to reformulate the model. In extreme cases, when the discrepancy cannot be remedied with minor modification, the theory embodied in the model can be considered refuted (Coltheart et al.). We start here by considering the effects of Neighbourhood Density since the present paper is, initially, a follow-up to a series of nine simulations with the DRC model reported by Reynolds and Besner (2002).

The Effect of Neighbourhood Density

An important question in reading research concerns how lexical knowledge affects the reading of new letter strings that could be words but are not. The present paper addresses this question by examining, both experimentally and computationally, how lexical knowledge affects nonword reading. Neighbourhood Density, the number of words that can be created from a stimulus by replacing a single letter at a time (Coltheart, Davelaar, Jonasson, & Besner, 1977), is used here to assess the contribution of lexical knowledge to nonword naming. Numerous studies have demonstrated that higher levels of N are associated with faster decisions for words and slower decisions for nonwords in the context of lexical decision (Andrews, 1997; Coltheart et al. …

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.