Exploring Unexpected Bilingualism in Autism: Enhanced Sensitivity to Non-Adjacent Dependencies

Abstract

Statistical learning refers to the ability to detect regularities from sensory input, including speech. Statistical learning plays a key role in language acquisition, particularly for complex structures, such as nonadjacent dependencies, that are ubiquitous in natural language syntax. This study investigates nonadjacent dependency learning in autistic children who acquire English through screen exposure, a phenomenon known as Unexpected Bilingualism (UB). Unlike their non-autistic peers, autistic-UB children acquire foreign languages with little interactional support. We hypothesize that this intensive experience with linguistic input should be associated in autistic-UB children with enhanced sensitivity to nonadjacent dependencies. An artificial language learning experiment confirmed that both non-autistic and autistic children with close to typical language ranges can learn non-adjacent dependencies from passive exposure to unfamiliar linguistic input. Crucially, autistic-UB exhibited significantly faster learning as compared to their autistic and non-autistic peers. This study documents that UB in autism is associated with distinct cognitive abilities.