Mécanismes d’apprentissage statistique et langage oral : de la théorie à la clinique

Background. Statistical learning (SL) mechanisms play a crucial role in oral language development. This field of research explores how individuals detect and use statistical regularities in linguistic stimuli to acquire language skills. Understanding these mechanisms provides valuable insights on language development and acquisition difficulties. However, SL and its clinical implications are still largely unknown to speech-language pathologists (SLPs). Aims. The aim of this literature review is to synthesize current knowledge on SL in oral language development, both typical and atypical, and to explore its possible applications in speech-language therapy. It aims to clarify the fundamental concepts of SL, analyze its role in language trajectories, and identify approaches for integrating these principles into therapeutic practices. Drawing on available scientific evidence, it also discusses methodological limitations and proposes perspectives for strengthening the links between theoretical research and clinic. Methods. This narrative review, with a heuristic and synthetic aim, was conducted based on a non-systematic but structured examination of the literature. Articles were selected using databases (PsycInfo, ERIC, MEDLINE, MLA, PubMed, and Google Scholar) with predefined keywords. Inclusion criteria focused on the field of study, population, and language. Priority was given to meta-analyses, systematic reviews, and experimental studies, incorporating both recent publications and key reference works. Results. Studies reveal the effectiveness of SL from early childhood in various aspects of language development. However, individual variations in SL, particularly in children with language difficulties, highlight the complexity of learning mechanisms. The article highlights the theoretical and methodological challenges in measuring and interpreting SL, as well as the practical implications for speech and language pathology intervention, proposing SL-based principles for improving therapeutic effectiveness. Conclusions. Statistical learning, with its powerful and rapid mechanisms, offers an interesting potential to optimize speech-language therapy interventions, complementing explicit learning. By focusing therapies on input and exploiting AS principles, SLPs can promote effective, effortless learning and generalization. Although further research is needed, current data encourage its integration into clinical practice, while opening the prospect of developing tools for screening and early intervention in various language domains.