Language Development in Context: A Longitudinal Study of Typically-Developing Children and Children with ASD

Background:  Children with Autism Spectrum Disorder (ASD) often display distinctive language development trajectories (Tek et al., 2013). Because language-learning is a social endeavor, these trajectories could be partially grounded in the dynamics that characterize the children's social and linguistic interactions (Waurlamont et al., 2014).

Objectives:  

We investigate language development trajectories and interpersonal linguistic adaptation over time in a longitudinal corpus of parent-child interactions.

Methods:  

Participants included 66 children, 33 with ASD (MA=33 months at visit 1) and 33 TD (MA=20 months at visit 1), matched on expressive language at Visit 1. Parent-child dyads engaged in 30 minutes of semi-structured naturalistic interaction at 6 visits, collected at 4 month intervals. We first quantified amount (number of word tokens), and complexity (number of word types) of the participants’ lexical behavior over time. Longitudinal adaptation was defined as the amount of variance a child’s/parent’s behavior at visit N explained in the parent’s/child’s behavior at visit N+L (baselined for her/his behavior at visit N), where L expressed an increasing distance between visits. We used mixed-effects growth curve models. The models included Mullen (Mullen 1995) and ADOS (Gotham 2009) scores as fixed factors.

Results:  

Our models described the developmental trajectories combining linear and quadratic trends (0.50 < R2m < 0.62, p<0.0001). For all measures we found main effects of time (linear β: -160, -39; quadratic: 22.92, 5.32), group (β: 332, 100.16), severity (β: -27, -8) and Mullen (β: -6, -1.6). Time significantly interacts with Mullen scores (linear β: 7.4, 1.92; quadratic β: -0.84, -0.2). Children with ASD showed shallower trajectories; higher Mullen scores were associated with higher intercepts and steeper trajectories, and higher ADOS scores were associated with lower intercepts.

Our models described longitudinal adaptation between consecutive visits for the number of word types and tokens (0.37 < R2m < 0.76, p<0.0001). Children adapted to parents (β: 0.04), but ADOS and Mullen scores modulated this. High ADOS scores implied less adaptation (β: -8, -2.3), as did low Mullen scores (β: 0.7, 2). Parents adapted to children (β: 0.15, 0.2) with no significant impact of group, ADOS, Mullen scores or age. Increasing the lag between visit (e.g. trying to predict the child’s behavior after 2 visits) decreased effect sizes and significance, with no statistical impact with a lag above 3 visits.

Conclusions:  

We investigated a model in which children and parents might mutually influence each other’s linguistic behavior over time, and asked whether this mechanism differed in children with ASD and their parents. Our results suggest that a quantifiable feedback loop between parents and children does exist in language development, and that this feedback loop is affected by the severity of autistic traits. This mutual adaptation mechanism appears to be in place in interactions between parents and children with ASD, though less strongly than in interactions with TD children. This may partially account for the shallower, but still present linguistic trajectories observed in children with ASD. Future work will assess more contingent dynamics of lexical and syntactic alignment and their role in language development.