Children with developmental language disorder (DLD) have severe difficulty with the acquisition of language. One hypothesis states that these children have a deficit in the ability of statistical learning, a domain-general learning mechanism that is important for extracting patterns and regularities from input implicitly. This dissertation focuses on the potential link between such a deficit in statistical learning and the lexical-semantic difficulties that are experienced by children with DLD. The following research questions are addressed: 1) do children with DLD have a deficit in various statistical learning abilities (word segmentation; cross-situational word learning and semantic categorization) compared to typically developing children and 2) are these statistical learning abilities related to lexical-semantic knowledge in children with DLD? Another research aim was to investigate ways of measuring statistical learning on-line in children with and without DLD.
This dissertation covers four empirical studies. We found evidence for a deficit in cross-situational word learning in children with DLD, indicating that they have more difficulty than typically developing children with learning to couple words to their referents in situations with referential ambiguity. We did not find evidence for or against a deficit in the other types of statistical learning ability in children with DLD, nor did we find evidence for or against a relationship between statistical learning and lexical-semantic knowledge. Concerning our on-line measures of statistical learning, using eye-gaze data as an index of cross-situational word learning shows promise.