Variability in the language input to children enhances learning in a treatment context

Artificial language learning studies have demonstrated that learners exposed to many different nonword combinations representing a grammatical form demonstrate rapid learning of that form without explicit instruction. However, learners presented with few exemplars, even when they are repeated frequently, fail to learn the underlying grammar. This study translated this experimental finding in a therapeutic context. Eighteen preschool children with language impairment received conversational recast treatment for morpheme errors. Over a 6-week period, half heard 12 unique verbs twice each during recasts (low-variability condition), and half heard 24 unique verbs (high-variability condition). Children's use of trained and untrained morphemes on generalization probes as well as spontaneous use of trained morphemes was tracked throughout treatment. The high-variability condition only produced significant change in children's use of trained morphemes, but not untrained morphemes. Data from individual children confirmed that more children in the high- than the low-variability condition showed a strong treatment effect. Children in the high-variability condition also produced significantly more unique utterances containing their trained morpheme than children in the low-variability condition. The results support the use of highly variable input in a therapeutic context to facilitate grammatical morpheme learning.