The Uses of Conversational Speech in Measuring Language Performance and Predicting Behavioural and Emotional Problems

Abstract

<p> Challenges to the diagnostic accuracy of standardized tests of language can make the utility of these measures on their own, problematic. Consequently, this research program uses tools of conversational analysis to study the speech of preschoolers and young adults.</p> <p> In the first of three studies we examine, from a purely data-driven approach, how conversational measures relate to one another and compare with WPPSI-III expressive and receptive vocabulary scores in assessing preschoolers' language. Mean length of utterance (MLU) was found to be the only conversation measure strongly related to WPPSI-III language scores. However, other conversation measures constituted reasonably stable factors that may have utility for children's language assessment.</p> <p> The second study uses the same sample of children to investigate what features of language best predict behavioural and emotional problems and whether conversation measures provide better prediction of these symptoms than standardized scores. Results indicated that conversation measures of language significantly improved prediction of Child Behavior Checklist (CBCL C-TRF) DSM-oriented and syndrome scales beyond that accounted for by WPPSI GLC scores.</p> <p> Finally, the third study uses conversational analysis to study the role of disfluencies in the speech of young adults with and without autism spectrum disorders (ASDs) to determine whether these features of speech serve listener or speaker-oriented functions. Individuals with ASD were observed to produce fewer filled pause words (ums and uhs) and revisions than controls, but more silent pauses. Filled-pause words, therefore, appear to be listener-oriented features of speech.</p> <p> Taken together, findings of this program of research highlight the importance of using conversational analysis as an alternative or in addition to standardized tests of language as well as inform what specific measures of language are best suited for this purpose.</p>