A Distributed, Developmental Model of Word Recognition and Naming
Abstract
A parallel distributed processing model of visual word recognition and pronunciation and of the acquisition of these skills is described. The model consists of a set of orthographic units used to code letter strings, a set of hidden units, and a set of phonemic units. Weights on connections between units were modified during a training phase using the back-propagation learning algorithm. The model takes letter strings as input and yields two types of output: a pattern of activation across the phonemic units, and a recreation of the input spelling pattern across the orthographic units. The model was trained on a corpus of 2897 English words that included most of the uninflected monosyllabic words in the language. The model provides detailed accounts of performance on two tasks, naming aloud and lexical decision, and simulates many aspects of human performance, including (a) differences between words in terms of processing difficulty; (b) pronunciation of novel items; (c) differences between readers in terms of word recognition skill; (d) transitions from beginning to skilled reading; and (e) differences in performance on the two tasks. The model's behavior early in the learning phase corresponds to that of children acquiring word recognition skills. Training with a smaller number of hidden units produces output characteristic of many poor readers. (kr)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 14, 1989
- Accession Number
- ADA218930
Entities
People
- James McClelland
- Mark S. Seidenberg
Organizations
- Carnegie Mellon University