L'acquisition de noms chez les enfants porteurs de dysphasie

Specific Language Impairment (SLI) is a neurodevelopmental disorder that affects the language development of children with a normal nonverbal intelligence and no history of neurological disorder nor auditory deficit (Leonard, 2014). A difficulty linked to SLI is the poor language productivity and the input dependency that children display compared with their peers: they tend to use a limited variety of verbal forms compared to younger siblings (Conti-Ramsden & Jones, 1997) and use a high proportions of the same verbs as their mothers (Jones & Conti-Ramsden, 1997). Recent studies suggested that these disorders could be linked to a deficit in generalizing constructions (constructions being units of language that vary in complexity and abstractness, Tomasello, 2009) from the input (Leroy et al., 2013). Specifically, some authors proposed that children with SLI would need more exemplars to abstract a given construction from the input and apply it to new instances (Leroy et al., 2013). In this study, we want to broaden this assumption to the acquisition of words: we evaluated the ability of children with SLI to generalize novel words, and in particular relational words. SLI seem indeed to be linked to an impairment in the processing of relations: children with SLI have poorer performance than their age-matched peers in linguistic or non-linguistic analogical tasks, which require to reason about relational structures (Leroy et al., 2014). We draw the hypothesis that children with SLI would need more exemplars in order to generalize words compared to age-matched but also to language-matched peers. We created a novel word extension task to improve this hypothesis: the first part is composed with nine test trials, each associated with a non-word. The child sees one, two, or three exemplars of a novel noun in short animations. The examiner tells him: “This is a /dyfɑ̃/”. Then, the child sees three other animations and the examiner says: “Show me which one of these is a /dyfɑ̃/”. The child has to choose the object that plays the same role as the exemplars among two distractors: an object that looks like one of the exemplar and a non-related distractor. The second part of the task is based on the same protocol but the exemplars are objects that have a specific spatial configuration (like symmetry). Here again, the child sees one, two, or three exemplars of a novel noun associated with the objects. The examiner tells him: “This is a /fegɑ̃/”. Then, the child sees three other objects and the examiner says: “Show me which one of these is a /fegɑ̃/”. The distractors are an object that has the same global form as one of the exemplar but without having the specific configuration required, and the non-related distractor. The participants are 20 children diagnosed with SLI aged from 6 to 12 years-old. Each child is match to a control child in age and non-verbal IQ on the one hand, and to a control child in linguistic age (based on a measure of vocabulary comprehension). Results show that children are better able to extend relational words when presented with several referents of these words, which confirms what has already been found in other experiments (Gentner, 2005). When the feature to be used is a spatial relation between the parts of the objects, children with SLI have more difficulties than age-matched controls extending new words, which confirms that SLI is linked to an impairment in the processing of relations and a greater dependance on perceptual information (Leroy et al., 2014). They also tend to benefit more than age-matched children from the presentation of several referents of a new word in order to extend it, what would mean that they greatly benefit from comparison and variability in order to identify relational similarities. It would be interesting to see if children can learn from several items that relational feature can be pertinent to define categories and extend words, as young children without SLI can do it with shape while children with SLI cannot (Collinson et al., 2015)