An Exploration of Methods for Rating Children's Productions of Sibilant Fricatives.

This paper examines three methods for providing ratings of within-category detail in children's productions of /s/ and /ʃ/. A group of listeners (n=61) participated in a rating task in which a forced-choice phoneme identification task was followed by one of three measures of phoneme goodness: visual analog scaling, direct magnitude estimation, or a Likert scale judgment. All three types of ratings were similarly correlated with sounds' acoustic characteristics. Visual analog scaling and Likert scale judgments had higher intra-rater reliability than did direct magnitude estimation. Moreover, both of them elicited a wider range of judgments than did direct magnitude estimation. Based on our evaluation, Likert scale judgments and visual analog scaling are equally useful tasks for eliciting within-category judgments. Of these two, visual analog scaling may be preferable because it allows for more distinct levels of response.