Lexical acquisition, generalization and Bayesian inference in children with Developmental Language Disorders

Developmental Language Disorders (DLD) are characterized among others by word learning impairment. The mechanisms underlying these difficulties remain misunderstood. Bayesian theories of cognition offer an interesting approach to study this phenomenon. It establishes inductive inference as a chore component of learning processes, based on the interaction of prior knowledge and environmental data. This study aims to determine if children with DLD can make inferences at two levels of abstraction, specific knowledge and rule acquisition, when prior knowledge is controlled, and if they can use prior knowledge as their peers do. Twenty children with DLD and twenty age-matched typically-developing children (6 to 12 year-old) were exposed to a new word learning task. Children are taught new categories of insects determined by particular physical characteristics that they have to infer. In a generalization task (second level of inference), they were presented with insects from unfamiliar categories. Finally, children are taught how physical properties relate to housing (baseline for prior knowledge) in a learning association task, and have to link each known insect with his housing. We assume that children with DLD would need more exposition to acquire similar bias in word learning and perform lower in the generalization task than their peers.