Comparison of Conversation and Narrative Language Abilities in Children with Autism: Diagnostic and Therapeutic Considerations

Abstract

Many children with autism experience significant difficulty with comprehension and expression of spoken language. Speech-language pathologists (SLPs) require sensitive assessments to describe the unique profiles of individual clients with autism so that appropriate treatments can be prescribed. Language sample analysis is a current best-practice for describing language profiles in children with language difficulties, though there is a limited literature describing clinical best-practices for children with autism. The purpose of this study was to analyze the difference in conversation and narrative language abilities in children with autism. Language ability was also analyzed in comparison to age-matched typically developing peers. This goal of this analysis was to assist in determining the context that reveals language deficits in children with autism to be utilized for diagnostic considerations and planning for intervention. Three children with autism, aged 6, participated in this study. Children were observed having a conversation with their parent or SLP and retelling a story from a wordless picture book originally told by their parent or SLP. Both samples were analyzed for grammatical and semantic errors. The conversation sample was analyzed for contingent responses, while narrative samples were analyzed for story grammar elements and organization. The analysis was completed using the Systematic Analysis of Language Transcripts (SALT) database. The microstructural analysis revealed that the participants’ relative weaknesses surrounded the complexity and accuracy of their utterances. The participants had difficulty producing syntactically accurate conversation samples demonstrated by a high number of word and utterance level errors. Within the narrative samples, the children produced utterances with relatively simple grammar and syntax demonstrated by very low MLU values compared to their peers. The macrostructural analysis of conversation samples revealed that 2 of the 3 participants produced half contingent and half non-contingent responses. Of those contingent responses, the majority were responses to wh-questions, despite one participant whose utterances were primarily self-expansions. The macrostructural analysis of narrative samples using the Narrative Scoring Scheme (NSS) revealed that the participants were at floor. Two participants achieved minimal scores for an introduction while the other participant received a minimum score for a conclusion. All narratives lacked cohesion and story grammar elements. Overall, the participants in this study did relatively worse in the narrative sample possibly due to the fact that it requires a higher level of language. Further research is needed to generalize these findings.