Speech Perception in Early Childhood: Contending with Speaker and Accent Variation

Understanding spoken language is much more complex than common intuition may suggest. Speaker-related variation in the realization of words due to changes in speaker gender and accent results in myriad possible realizations of the same word. Although adult listeners can easily accommodate such speaker-specific pronunciations, research suggests that children have more difficulty with this variability. In this thesis, I examine when and how children learn to contend with speaker and accent variation. In three lines of research, I test the hypothesis that infants and toddlers may be less challenged by acoustic variability than past research suggests. In Chapter 2, I show that 15-month-old infants, like adults, can readily accommodate an unfamiliar Australian accent after hearing a 2-minute story produced in Australian English. Interestingly, speaker exposure only elicits accommodation when infants are highly familiar with the words in the story. These findings are consistent with the hypothesis that lexical feedback serves as the mechanism inducing accent adaptation. More support for this idea comes from Chapter 3, where even 20-month-olds were found to experience difficulty accommodating a speaker’s unfamiliar accent when the speaker uses hard words that are not consistently known by children this age. Chapter 3 further shows that 28-month-olds, but not 20-month-olds, recognize words produced in sentence frames by a speaker with an unfamiliar accent without any prior exposure to the accent, thereby providing converging evidence that young children are proficient at dealing with accent variability. Chapter 4 subsequently demonstrates that even 7.5-month-olds can contend with some forms of speaker variation when tested under relatively natural listening conditions. Taken together, these three lines of research provide compelling evidence for the view that early speech processing abilities are remarkably sophisticated and that the transition from infant to adult listener may be less abrupt than currently thought.Ph