Conversational Dynamics in a Longitudinal Corpus of Caregiver-Child Interactions

Background:  Conversational turn-taking provides a scaffold that helps children learn the elements and rules of language (Dale & Spivey 2006, Yurofsky et al 2016). This relies heavily on social feedback and engagement (Fusaroli et al 2014), and thus conversations involving children with ASD are expected to show different turn-taking dynamics than conversations involving typically-developing (TD) children. These dynamics remain relatively unexplored.

Objectives:  Our goal was to explore the turn-taking dynamics between caregivers and children with ASD. More specifically we asked: what significant differences can be observed in the number, duration, and time between conversational contributions by adult caregivers and children in dyads with ASD and TD children?

Methods:  We investigated spontaneous speech in caregiver-child dyads from a longitudinal corpus (6 visits over 2 years), consisting of 30 minutes of controlled playful activities between caregivers and 67 children: 35 with ASD (MA = 33 months) and 32 initially-language-matched typically developing (TD; MA = 20 months, cf. Goodwin et al. 2012). Time-coded transcripts were used to calculate utterance duration in seconds, interturn latency (seconds between utterances), and number, and timing of each interlocutor’s utterances. Mixed effects models were employed to assess the relation between diagnosis and these measures, separately in children and caregivers, controlling for participant ID, time within conversation, and visit.

Results:  While there was no main effect of diagnosis, the conversations displayed interactions between diagnosis and time. TD children increased their turn duration over the course of a conversation more than children with ASD (Beta=-0.05, SE=0.02, p=0.003), and this difference increased in later visits (Beta=0.01, SE<0.01, p=0.007). Opposite effects were displayed in interturn latency as the children in the ASD group increased their inter-turn latency more as conversations went on than did their TD peers (Beta>-0.01, SE<0.01, p=0.037). Adults showed no differences in inter-turn latency, but those with TD children increased the length of their turns more as the children aged than adults with children with ASD did (Beta: -0.06, SE=0.02, p=0.006). Moreover, the conversations all showed interpersonal adaptation: the greater the child’s latency (Beta=0.1, SE=0.02, p<0.001) and the longer their utterance duration (Beta=1.7, SE=0.01, p<0.001), the more turns per unit of time the adults took.

Conclusions:  We used turn duration, time between turns, and number of turns per unit of time to probe conversational dynamics between adult caregivers and children with ASD. Although preliminary, our results suggest that the conversational environments in which children learn and practice their linguistic skills are also affected by ASD. Not only did the conversational contributions of children with ASD differ from those of their TD peers, the contributions of adult caregivers also differed between the group, presumably as adults adapt their conversational turns to the child they are speaking with. This highlights the importance of investigating language acquisition in children with ASD within the context of dyadic interaction.