Purpose: Estimating the sequential associations between educators' and children's talk during language learning interactions requires careful consideration of factors that may impact measurement stability and resultant inferences. This research note will describe a preliminary study that used generalizability theory to understand the contribution of two measurement conditions-occasions and raters-on estimates of sequential associations between educator talk and autistic preschooler talk in inclusive preschool classrooms.
Method: We used an existing data set of four 15-min video-recorded occasions of educator-child interactions for 11 autistic preschoolers during free-play in their inclusive classroom. Two trained raters coded all videos for preschooler talk and type of educator talk (i.e., opportunities for expressive language [OELs], other talk). We conducted two generalizability studies on sequential association estimates for two interaction directions (i.e., preschooler talk following educator OEL and educator talk following preschooler talk). We conducted a series of decision studies to explore configurations of measurement conditions to optimize future investigations.
Results: We had unstable estimates for both interaction directions in our current methodological approach, with raters accounting for minimal error and occasions accounting for considerable error. Future investigations would require at least six observation occasions for stable estimates of the sequential association between autistic preschooler talk following educator OEL that was stable after six occasions. More than 15 occasions were required for stable estimates of the association between educator talk following autistic preschooler talk.
Conclusion: We will share recommendations and implications for future investigations to estimate educator and child talk sequential associations within preschool language interactions.