Models for Predicting How Adults Pronounce Vowel Digraph Spellings in Unfamiliar Words

Abstract

This study explored the relationships between type and token frequencies and contextual position effects in pronunciation of written English. Specifically, the major question was whether or not vowel cluster pronunciation preferences of adult readers were more affected by frequency of occurrence than by graphemic environment. Two opposing hypotheses were tested regarding four vowel diagraph spellings. Six synthetic words were constructed for each vowel cluster according to contextual and word position constraints. The subjects were 51 undergraduates whose task was to read the synthetic words and note how they pronounced the underlined vowel cluster. Three models were constructed to assess the hypotheses and to predict responses for each vowel cluster. The models were a final consonant model, a variant type-token model, and an invariant principal response model. Several data analysis techniques were used. The final consonant model was superior to the other two models, but it was found that other factors, not yet assessed, were present in the results.