Early Speech and Language Assessment in Toddlers with Autism

Background: Autism is a complex, pervasive developmental disorder with significant impairments in social and communication development as well as repetitive behaviors.  There is evidence that early intensive behavioral intervention initiated at preschool age can result in substantial improvements in a large subset of children with autism, including significant gains in IQ, language, and educational placement. Early intervention has been shown to demonstrate gains for children with autism, but some children respond more to intervention than others. The factors that determine responsiveness to intervention are not well understood.  However, some evidence suggests that early speech and language abilities are likely to be significant predictors of response to early intervention. 

Objectives: In this study, we examine a novel measure of speech production, referred to as “speech articulatory signatures,” and its relation to other aspects of language ability in young children with ASD who are entering into a randomized clinical trial of early intervention.  Such speech signatures, which reflect the maturation of fine motor control in speech, have been shown to follow a specific developmental time course in typically developing children (Singh L & Singh NC, 2008, Developmental Science, 11: 467-73). 

Methods: The speech signature measure uses modern spectral analysis to investigate the development of articulatory and phonological features in vocal utterances: syllabic rhythm, format transition, and place of articulation. The current sample consists of 48 toddlers diagnosed with ASD (M CA = 23.48 mos., SD = 3.92, range 18-30 mos., 37 M, 11F), as well as chronological age matched children with typical development, and chronological age and mental age matched children with idiopathic developmental delay.  The children with ASD are participating in the NIMH-funded UW Early STAART study, which uses an approach that involves a relationship-based intervention framework to accomplish developmentally-based objectives using naturalistic application of applied behavior analytic principles in children diagnosed before 30 months.  Speech samples were gathered during diagnostic and other free play conditions video- and audio-taped for later analysis.  Other measures available include the language subscales on the Mullen Scales of Early Learning and the Vineland Adaptive Behavior Scales, social orienting, and joint attention. 

Results:   Initial spectral analyses of speech samples gathered in naturalistic conditions were found to yield reliable speech signature data in all three comparison groups.

Conclusions:   Reliable measures of articulatory and phonological features of speech can be obtained from naturalistic speech samples taken in free play situations.  Results of correlational analyses examining the association between characteristics of speech signatures and concurrent language ability will be reported.  In addition, we will examine whether (1) toddlers with ASD differ from those with typical development and developmental delay in terms of their early speech signatures, and (2) whether, among the children with ASD, speech signature measures are correlated with other measures of language and social attention, all of which were assessed before entering the intervention study (Time 1).  In the future, we plan to examine how the speech signature measure predicts response to treatment and whether it is a better predictor than standard measures of language.