Vowel dominance in overregularizations

When children produce regularizations like comed, not all verbs are equally likely to be regularized. Several variables (e.g. lexical frequency) have been shown to be relevant, but not all the variability between verbs is understood. It is argued here that one predictor is which vowels are present in the base form vs. the past tense form. Using a notion of recessive vs. dominant vowel (where recessive vowels are more likely to be replaced by dominant vowels than vice versa) based on adult phonological processing, it is predicted that regularizations should be likely when the base vowel is dominant and unlikely when the past tense vowel is dominant. Data from 17 children reported in the literature, aged 1;6-5;6, show that this prediction is correct. Implications for the role of phonological variables in the processing of irregular past tense forms are discussed.