Categorization and word extension in children with Developmental Language Disorder

Background: Word learning difficulties are often part of the deficits presented by children with Developmental Language Disorders (DLD). Word learning is known to be an inference-based processed, constrained by categorization, which helps the extension of new words to unfamiliar entities. These processes appear integrated in Bayesian models of cognition, which offer an interesting approach to explore this phenomenon. Indeed, learning takes place thanks to an inductive inference process that recruits prior knowledge and principles of statistical learning (detection of regularities). Our study aims at
(1) defining if children with DLD can draw inductive inferences in a word-learning context using categorization;
(2) defining the impact of prior knowledge on categorization tasks, i.e. defining if the task is easier when prior knowledge is available vs not available and exploring the strategies used by children to categorize unfamiliar stimuli.
Methods: Thirteen children with DLD (between 6;0 and 9;0) and fifteen age-matched controls were exposed to a word-learning task in which they faced with exemplars of objects associated with pseudo-words. Objects belong to three categories spreading across three hierarchical levels. For each item, children chose among a set of test objects from the same categories which one(s) could be labelled the same (word extension). Our design included two parts: familiar and unfamiliar categories.
Results and conclusions: Our results replicated previous work on familiar categories, which showed similar performances for children with and without DLD in a word extension task. However, differences between groups appeared for unfamiliar categories, suggesting difficulty in category-based inference and word extension when prior knowledge is not available to support the inference process of children with DLD.