Discovering large grain sizes in a transparent orthography: Insights from a connectionist model of Italian word naming

Abstract

Classic connectionist models of reading have traditionally focused on English, a language with a quasiregular (deep) relationship between orthography and phonology, and very little work has been conducted on more transparent (shallow) orthographies. This paper introduces a parallel distributed processing (PDP) model of reading for Italian. The model was explicitly developed in order to deal with polysyllabic words and stress assignment. One of the core issues regarding such PDP models is whether they can show sensitivity to large grain sizes, as documented by the existence of morphological and neighbourhood effects in nonword reading aloud showed by native Italian speakers (Arduino & Burani, 2004; Burani, Marcolini, de Luca, & Zoccolotti, 2008). The model is successful in simulating these effects, previously accounted for by dual route architectures. The model was also able to simulate stress consistency effects.

Keywords:

• Grain size
• Modelling
• Morphology
• Naming
• Neighbourhood size
• Orthographic depth
• Stress assignment

Acknowledgements

This work was supported by an EU Sixth Framework Marie Curie Research Training Network Program in Language and Brain (http://www.ynic.york.ac.uk/rtn-lab). PM was supported by ESRC Grant RES-000-22-1975.

We are indebted to Jangfeng Yang for invaluable support provided during the development of the model, and to Cristina Burani, Jelena Mirkovic, Jason Zevin, and two anonymous reviewers for helpful comments on an earlier version of the draft. An executable version of the model can be downloaded from: http://www.lancs.ac.uk/staff/monaghan/resources/italianmodel