par Dumont, Charlotte
Président du jury Janssens, Sonja
Promoteur Kissine, Mikhail
Co-Promoteur Destrebecqz, Arnaud
Publication Non publié, 2025-05-07
Thèse de doctorat
Résumé : Typical language acquisition is bolstered by face-to-face socio-communicative interactions. Marked challenges in socio-communicative abilities are a core characteristic of autism and likely contribute to the frequent language disorders and delays in autistic children. And, indeed, autistic children with reduced joint attention abilities are more likely to remain minimally verbal. However, around 70% of autistic individuals eventually acquire structural language skills despite persistent socio-communicative difficulties. In this thesis, I explore the role of non-socially mediated mechanisms that may explain language variability in autistic individuals, challenging the focus on socio-communicative abilities. This is done in two ways.First, I investigate the potential role of statistical learning, the mechanism underlying the automatic extraction of regularities in the environment, which likely plays a central role in language acquisition, typical or not. Current research on statistical learning and autism mainly focuses on adjacent dependencies in verbal and older autistic individuals, neglecting both more complex structures and less verbal autistic individuals. Focusing solely on adjacent dependencies, which are associated with early milestones of language acquisition such as word segmentation, may not reveal meaningful variability in verbal autistic individuals. While non- or minimally verbal autistic individuals often acquire a small repertoire of isolated words and sometimes achieve a few two-word combinations, they do not display productive and functional use of syntax. Sensitivity to nonadjacent dependencies is a crucial aspect of natural language syntax. These dependencies have an AXB structure, where A and B are related, but separated by an independent and variable sequence X. Differences in sensitivity to such long-distance dependencies may then help delineate different linguistic profiles in autism. For example, non- or minimally verbal children may exhibit reduced sensitivity to nonadjacent dependencies, potentially accounting for their challenges to develop complex syntactic constructions as compared to their verbal autistic and non-autistic peers.Second, I study the phenomenon of Unexpected Bilingualism in autistic children. Autistic unexpected bilinguals achieve advanced mastery of a first or second language that is not communicatively used around them and do so exclusively through non-socially mediated screen exposure. Although there are handful case descriptions of Unexpected Bilingualism in autism, no study to date has examined the mechanisms that could underlie such an atypical mode of language acquisition. As Unexpected Bilingualism emerges from exposure to screens that involve no socio-communicative interaction, its existence challenges the assumption that socio-communicative skills are necessary for language development. The phenomenon of Unexpected Bilingualism raises the possibility that some autistic children rely more on the internal structure of the language input rather than on its socio-communicative function, suggesting the existence of an alternative path to language in these children. Chapter 1 addresses the first line of research, exploring statistical learning as a compensatory mechanism underlying language acquisition in young autistic children. I first conducted four pilot studies to refine the methodology for assessing statistical learning in preschoolers. These studies are fully detailed in the Coda of the dissertation, where I also discuss the challenges inherent in studying statistical learning in children who are too old for experiments used with infant paradigms but too young for paradigms that rely on explicit instructions and responses. However, given the limited success of these pilot studies, in Chapter 1, I refocus on autistic aged 5 to 8-year-olds, whose linguistic profiles range from fully verbal to nonverbal. I examine nonadjacent dependency learning as a critical mechanism for morpho-syntactic abilities and compare its predictive power to that of joint attention. Neither joint attention nor nonadjacent dependency learning predicted vocabulary or morphosyntactic skills in autistic children, once non-verbal cognitive abilities are factored in. That said, only a small proportion of autistic children completed the nonadjacent dependency task, with non-verbal cognitive abilities also predicting task completion. In Chapters 2 and 3 I focus on a subgroup of autistic pre-adolescents and adolescents with an Unexpected Bilingualism profile. Among a sample of 48 autistic participants, 25% acquired English only through early exposure to screens. I test the hypothesis that autistic children with an Unexpected Bilingualism profile rely more on structural properties of linguistic input, such as statistical regularities and prosodic features of speech, such as pitch, rather than social cues, to acquire language. In the experiment reported in Chapter 2, autistic Unexpected Bilingual children displayed faster learning of nonadjacent dependencies than their autistic and non-autistic peers. These findings suggest an enhanced sensitivity to statistical regularities in Unexpected Bilingual autistic children. In Chapter 3, I administered a pitch discrimination task, as pitch is an important acoustic feature and cue for the acquisition of phonological representations. This study revealed better pitch discrimination in Unexpected Bilinguals autistic children compared to their autistic and non-autistic peers. Results from Chapter 1 underscore the significant methodological and practical challenges of collecting behavioral data from young autistic participants with diverse language profiles. While the hypothesis that statistical learning underpins language acquisition difficulties in autism is compelling, it remains difficult to validate empirically, particularly in younger populations and when using behavioral paradigms. The findings from Chapters 2 and 3 suggest that individuals with Unexpected Bilingualism have a distinct cognitive profile that may drive their atypical language acquisition trajectory. The high prevalence of Unexpected Bilingualism in my autistic sample (25%) also strongly suggests that language acquisition may follow a different path in autism and that Unexpected Bilingualism is a promising avenue for better understanding language acquisition in autism.

Documents en relation