A developmental model of vocabulary production

Fedor, A. 1 , Best, W. 2 , Masterson, J. 3 & Thomas, M. 1

1 Department of Psychological Sciences, Birkbeck College, London, UK
2 Division of Psychology and Language Sciences, University College London, UK
3 Department of Psychology and Human Development, Institute of Education, University of London, UK

We present a developmental model of vocabulary production, which provides a formal specification of the developmental mechanisms that drive naming abilities. The key aims are to explain the typical development of naming ability, the associated developmental disorders, and to predict possible strategies for intervention.
Children’s productive vocabulary development is characterized by different profiles of naming accuracy, latency and error types (mainly semantic, phonetic and mixed errors). Deviations from typical development can be the result of different impairments, leading respectively to semantic difficulties, problems with accessing the word form, phonological difficulties, or multi-loci word retrieval problems. On current theoretical accounts, the localisation, degree and the nature of impairments identify subgroups of children with word-finding difficulties (see Messer & Dockrell, 2006 for review).
Previous computational models of naming have demonstrated how associations can be learned between phonological and semantic representations. However, these representations have been either pre-specified or restricted to simulating normal development (e.g., Foygel & Dell, 2010; Li et al., 2007; Mayor & Plunkett, 2010; Plunkett et al., 1992). In our model, we allow semantic and phonological representations to themselves be the outcome of a developmental process. We then explore the consequences of these representations developing under atypical constraints. The model includes separate semantic and phonological components, thus enabling investigation of different deficits within these modules, as well as impairments occurring during the learning of associations between them. The model is calibrated against new empirical data from typically developing children and children with word-finding difficulties, focusing on accuracy, latency, and error patterns in naming tasks.
Lastly, we discuss the potential of models of atypical productive vocabulary development to predict optimal intervention approaches for children with different underlying difficulties. Here, the main question is whether interventions targeted to specific processing difficulties are more effective than those that treat the entire naming system.