The Role of Augmented Language Input on the Expressive Language Growth of Young Children with Autism Spectrum Disorder (ASD)

Background: The development of language in children with autism spectrum disorder (ASD) is a potent predictor of positive social outcomes (Anderson et al., 2009). Yet despite intensive early intervention approximately 30% of children fail to develop functional language by the time they enter school (NRC, 2001). One promising intervention is the use of speech generating devices (SGD), which have shown treatment effects when used in isolation (Van Der Meer & Rispoli, 2010) or in conjunction with a behavioral treatment (Kasari et al., 2014). There are a number of hypotheses as to the mechanism by which SGD lead to language growth such as the consistency of the language produced by SGD and the pairing of the language with a visual. There is little empirical evidence supporting these hypotheses (Blischak, Lombardino & Dyson, 2003).

Objectives: To test the moderating effect of the language input produced by SGD versus verbal language input from an adult on the expressive language growth of children with ASD during an intervention trial.

Methods: Thirty-one children were included with a mean age of 6.44 years (SD=1.23). All children received a 24-week behavioral intervention (JASPER+EMT) with an SGD used by the interventionist to model language (see Kasari et al., 2014 for intervention overview). The dependent variables were the number of spontaneous comments, word roots and requests measured using the Natural Language Sample. The independent variables were the proportion of children’s responses to adult language (verbal and/or SGD models). A response was defined as the child producing contingent language within five seconds of adult’s verbal or SGD language models.

Bi-variate correlations were run between the outcomes, children’s response to models, Leiter age equivalency and ADOS severity scores. Next, linear regression models were run for each dependent variable at exit with the proportion of responses to SGD and verbal models at entry entered as independent variables controlling for cognitive ability and entry scores on the dependent variables.

Results: The percentage of responses to language input produced by the SGD was correlated with the number of different word roots (r=.57; p<.01) and spontaneous requests (r=.47; p<.05) at entry. None of the outcomes were correlated with the proportion of responses to verbal models.

The regression equations for spontaneous comments and number of different word roots were significant, F (4, 17) = 5.28, p< .01 and F (4, 17) = 14.67, p< .01.

Proportion of response to SGD models at entry (but not verbal models) was predictive of the number of spontaneous comments and different word roots at exit after controlling for entry scores and cognitive ability, t(21) = 2.96, p < .01 and t(21) = 2.85, p = .01 respectively.

Conclusions: These data suggest that children process augmented language models differently than adult verbal models. It could be that those children with higher receptive language or cognitive abilities are able to benefit from the input from the devices (Sevcik, 2006). A direction for future research will be to test the role of these potential moderators with a larger sample of children.