Evaluating listener bias introduced by two speaker groups on VAS ratings: A methodological investigation

Abstract

Objectives: The goal of this experiment was to determine whether listeners rate /t/ and /k/ consonants produced by 3 to 5-year-old children with cochlear implants (CIs) differently when in a blocked condition that did not include productions by children with normal hearing (NH) compared to an unblocked condition which did include such productions. Blocking has been shown to influence results on categorical tasks. Some research suggests that judgments made using continuous a visual analog scale (VAS) are less susceptible to bias than judgments made using a categorical system. However, the research on VAS and bias is sparse, with no research on the interaction of VAS ratings and perceptual bias when bias is introduced to an experiment via a comparison of two groups. This study used word-initial /t/ and /k/ CV sequences produced by children with NH and children with CIs to investigate the effect of blocking on VAS judgments. Design: 48 adult participants were recruited at the University of Maryland. Each participant rated 500 CV tokens in a VAS experiment. Half of the participants were assigned to the blocked condition, and half to the unblocked condition. Mixed-effects models were used to analyze the ratings of the tokens produced by children with CIs to see if there was a significant change from the blocked to the unblocked condition. Results: For both the t-like and k-like tokens, models showed significant effects of intercept and transcription category. Ratings for production by children with CIs were not significantly different across the two conditions. Conclusions: The VAS ratings of tokens from children with CIs did not differ in the blocked and unblocked condition. This result supports the finding that VAS may be less susceptible to bias than categorical judgments. In future studies, researchers may choose blocked or unblocked designs to compare these two groups of speakers, depending on which design is better suited to answer individual research questions.