An Application of Generalizability Theory to Characterize Language Learning Interactions for Young Children with Autism in Their Inclusive Classroom
For children with autism spectrum disorder (ASD), features of the disorder can impact the nature, frequency, and length of adult-child interactions important for language learning (Carpenter & Tomasello, 2000; Landa, 2007; Wetherby et al., 2004). Given the important role that interactions play in language learning for young children, it is important to examine the means by which inferences may be made about children’s language learning environments. In other words, inferences about the nature, frequency, and length of interactions that produce clinical effects are based on researchers’ systematic observation and sampling of certain behaviors for certain periods of time within certain settings (Irvin, Boyd, & Odom, 2015; Irvin, Hume, Boyd, McBee, & Odom, 2013; Sanders et al., 2016; Yoder, Lloyd, & Symons, 2018). The sampling approach relative to different facets of measurement (e.g., observers, occasions, contexts) contribute variability that may have a significant, yet often overlooked, impact on the inferences that may be made by researchers when interpreting their results (Webb, Shavelson, & Haertel, 2006).
Guided by generalizability theory (g-theory; Cronbach, Rajaratnam, & Gleser, 1963; Shavelson & Webb, 2006), the study objectives are to (a) describe methods for characterizing language learning environments of young children with ASD based on relevant measurement facets and (b) provide empirical guidance for conditions of measurement that should guide future investigations to optimize both the reliability of inferences and efficiency of data collection.
Using systematic observation methods, video recordings of interactions between 10 children, ages 3- to 5-years-old with ASD, and adults in their inclusive preschool classrooms during exploratory play are being examined. Consistent with g-theory, two measurement facets were identified: occasions and observers. As such, 15-minute recordings for each participant are being gathered across four occasions within a three-week timespan. After ensuring 90% agreement across observers, each observer is coding each video for frequency and type of adult language behaviors (i.e.., open-ended questions, yes/no questions, choice questions, completion prompts, imitations, comments) and child communicative acts. The generalizability study is utilizing a person x occasion x observer random effects Analysis of Variance (ANOVA) model. A decision study will follow with levels of occasions and observers manipulated to determine the optimal conditions of measurement that lead to stable estimates of observed frequencies and types of adult behaviors and child communicative acts.
With the investigation underway, the researchers will present results of the generalizability study, displaying estimates of magnitude with which (a) variability in the person, (b) variability in the facets (i.e., occasions and observers), and (c) variability of the interaction of the person with the facets contribute to the observed measures of adult and child behaviors. Likewise, the researchers will present the results of the decision study, such that the impact of various levels of facets on error variance can be seen.
The researchers will provide guidance on measurement conditions that are most relevant for researchers in interested in sampling language behaviors in inclusive preschool environments and provide recommendations for incorporating g-theory into investigations utilizing systematic observation methods.