Learning bias of phonological alternation in children learning English

Abstract

It has been assumed that output-to-output faithfulness constraints are undominated in the initial stages of learning (McCarthy, 1998; Hayes, 2004), but they are gradually demoted in the hierarchy as children are increasingly exposed to alternating forms. A prediction ensuing from this claim is that children will initially favor non-alternating forms, but they will eventually produce alternating forms as adults do. This study aims to provide empirical evidence for this claim with the case of American English Flapping. According to Bernhardt and Stemberger (1998), a child learning English did not apply flapping rule where adults do, for the sake of paradigm leveling; for instance, she produced si[t]ing (rather than si[ɾ]ing) for sitting, on the model of si[t]. Using six children's longitudinal spontaneous speech recorded from the age of 0;11 to 4;0 (Demuth, Culbertson, & Alter, 2006), this study investigated how childrens probability of producing each variant of word-medial /t/ and /d/ changes as a function of age. The results demonstrated that the frequency of the variants that are more faithful to the base (unsuffixed) form, such as si[t]ing, decreases as a function of age, while the frequency of [ɾ] was found to increase over time. The findings indicate that children in fact are biased against alternation, especially in the initial stages of learning.