Application of statistical learning principles in Cantonese grammar intervention for preschool children : effects of exemplar variability

Abstract

Statistical learning has been shown to play a central role in child language acquisition, and its principles have been applied in grammar intervention for children with Developmental Language Disorder (DLD). One application is evident through Enhanced Conversation Recast (ECR), an intervention which involves manipulation of exemplar variability (variations in the non-essential elements of the construction) in traditional Conversational Recast. It was proposed that the additional manipulation increases the salience of the relation between two grammatical elements, facilitating statistical learning. Results from studies examining the efficacy of ECR showed that given the same dose number, high exemplar variability in the input consistently led to better learning outcomes than low exemplar variability in the learning of English grammatical morphemes. Studies of exemplar variability with natural language learning were conducted predominantly in English, and its effect has not been examined across language typologies. Aspect markers are grammatical morphemes that have consistently been reported to be problematic for Cantonese-speaking children with DLD. To contribute to the limited intervention evidence for statical learning in different languages, this study sought to examine whether the same variability effect applies to the learning of Cantonese aspect markers for Cantonese speaking children with DLD. The current study included 34 children aged between 3 to 5 (12 typically developing [TD] and 22 with weak oral language skills [WOL]) who participated in an individual training program that targeted Cantonese aspect markers. Focused stimulation was used alongside conversational recast in the intervention. The children received twelve 30-minute individual training sessions and were randomly assigned to either the high or the low exemplar variability condition. Both conditions had the same dosage, 24 doses, in each of the 12 sessions, resulting in a cumulative intervention intensity of 288 doses. There were 24 unique verbs in the high exemplar variability condition and 12 unique verbs that were repeated once in the low exemplar variability condition. During the session, each child engaged in various activities with a trained speech language pathology student that encouraged the production of aspect markers in a natural context with recast as feedback or focused stimulation as language input. High intervention fidelity and scoring reliability were observed for the intervention (>90%). Data analysis utilized generalized linear mixed effect modeling. Results revealed that children in both the high and the low exemplar variability conditions made statistically significant yet small gains with their target aspect markers. However, no statistically significant differences were observed between children in the two conditions. These results were inconsistent with reports from previous studies which favored the use of high exemplar variability language input. Potential limitations of the use of high exemplar variability, low performance in both variability conditions, and methodological issues are discussed in relation to the nature of Cantonese, particularly the grammatically optional nature of Cantonese aspect markers, and the intervention procedures used.