The Effects of Listener Bias on the Perception of Accuracy in Children’s Speech

Abstract

Phonetic transcription is a tool that has long been used by speech-language pathologists (SLPs) to record children's productions of speech sounds. Transcription employs a level of of detail that allows SLPs to assess whether children produce speech sounds accurately, and to track progress during the treatment of production errors. Transcription has generally been considered to be an objective method, and not subject to bias when used by clinicians with professional training. In contrast, previous work (Schellinger, Edwards, Munson, and Beckman, 2008a,b,c) suggested that listener's expectations about a child's age and development can affect his or her rating and subsequent transcription of the child's production. Schellinger et al. found only a small biasing effect in their experiment. The goal of this UROP project was to further investigate the strength of listener bias, and to explore the factors that influence bias. In the Spring 2009, I ran three experiments in which subjects rated the accuracy of 200 productions of the "s" sound by preschool children. Listeners were led to believe they were produced either by an older, typically developing child, or a younger child with speech-production errors. I examined whether the adult listeners shifted their criteria for a correct "s" as a function of their assumptions about the child's age/ability level. The three experiments differed slightly in the expectations about the child that were given to the subjects. This allowed us to better understand why listeners are biased by expectations. The results of the project will help us develop more objective clinical assessments.