Reliability of professional and naïve listeners on perceptual evaluation of different types of voice sample in children

Abstract

This study investigated the effects of speech contexts, severity of voice problems and professional background on reliability of perceptual voice evaluation in children. Two groups of listener were recruited. The first group comprised of 10 speech therapists with 1 to 13 years of experience in working with pediatric speech and language caseloads (Professional Group). The second group comprised of 20 naïve listeners (Naïve Group). Both groups of listeners were asked to rate perceptually the severity levels of voice samples of 40 children speakers on three vocal parameters (roughness, breathiness and overall severity). For each child, there were three types of voice samples including prolongation of vowel /a/, sentence and short passage. Intraclass correlation (ICC) was calculated to identify intra- and inter-rater reliability of perceptual voice evaluation by the two groups of listener. Results revealed higher rater reliabilities 1) in passage reading, and 2) by professional listeners. Interestingly, disagreements were noted between the two groups on reliability across voices with different severity levels. These findings contradicted the initial hypothesis of higher reliability on normal and severely dysphonic voices in children.