Abstract :
[en] Developmental language disorders are related to morphosyntactic difficulties and to a deficit in generalization (Riches, Faragher, & Conti-Rasmden, 2006). This study aims at examining the ability of children with DLD to generalize constructions, i.e. form and meaning pairings (Goldberg, 2003). Our first objective is to compare construction generalization in children with and without DLD. As a second objective, we want to evaluate the impact of a training input involving progressive alignment in comparison with an input with high variability. Progressive alignment (i.e. the comparison of close items followed by the comparison of far items) has indeed been found to facilitate the process of comparison and to enhance word extension in young children (Childers et al., 2016; Gentner, Anggoro, & Klibanoff, 2011; Kotovsky & Gentner, 1996). Finally, our last objective is to examine the impact of inhibition on construction generalization, as this function has already been linked to the abstraction of patterns and categories (Richland, Morrison, & Holyoak, 2006; Snape & Krott, 2018).
We replicated the protocol of Casenhiser and Goldberg (2005): children are presented with sentences created according to a novel construction (a Subject-Object-Verb structure linked to a meaning of apparition) which they have to generalize to new sentences in the testing phase. However, we modified the training conditions of Casenhiser and Goldberg (2005). In the first condition, all the sentences are different. In the second condition, the sentences are constructed according to the principle of progressive alignment: at the beginning they have lexical items in common but they progressively become totally different. Thirty children with DLD were recruited and matched to a control child according to a sentence comprehension measure. Participants were divided in two groups: one presented with the progressive alignment condition, and one presented with the high variability condition. The training phase consists in 8 videos picturing a character which appears on the screen, associated with a SOV sentence. The testing phase includes 12 SOV or transitive sentences for which children have to select the corresponding video, picturing a character appearing on the screen or performing an action.
Results show that children with DLD perform similarly as their language-matched peers and that both groups have better results in the progressive alignment condition. Moreover, the performance of both groups differs from chance level in the progressive alignment condition only. We highlight here the benefit of progressive alignment compared to high variability on construction generalization in children without language disorders, but also in children with DLD. This could lead to promising trails in conducting interventions with these children.
