Evaluating the Validity of Vocalization Measures for Assessing Vocal Development in Young Children with Autism Spectrum Disorder
This study evaluates evidence of validity of multiple vocal measures purported to assess vocal development in young children with autism spectrum disorder in the early stages of language learning. Participants include 87 children with autism spectrum disorder (M = 23.42 months, SD = 3.98 months). We evaluated evidence of convergent validity, divergent validity, sensitivity to change, and incremental validity of 14 vocal variables measured across a 12-month span. These variables included conventionally-coded and automated variables that are purported to assess volubility and vocal complexity. A conventionally-coded approach quantified communicative use of vocalizations. An automated approach quantified reciprocal vocal contingency. Incremental validity analyses compared the contributions of more elaborate or more costly vocal variables relative to less elaborate or less costly variables. Growth curve modeling was the primary analysis approach. After eliminating redundant variables, nine vocal variables remained. Because there is no agreed upon method for comparing validity across variables, results are discussed from multiple perspectives for comparing validity evidence among variables. Depending on the perspective taken, different sets of variables are highlighted as most valid. Across perspectives, the results highlight the value of measuring communicative use and complexity through conventional means. The varying levels of validity evidence for automated vocal variables should be considered when selecting among the three automated vocal variables. However, the validity of reciprocal vocal contingency was not supported after controlling for the automated volubility vocal variable, and one of the automated complexity variables, infraphonological vocal development score, had little validity evidence in this sample.