Predicting language and literacy growth in children with ASD using statistical learning - Abstract
Children with Autism Spectrum Disorder (ASD) show enormous heterogeneity in core language abilities, such
as phonology and grammar. Over 50% of verbal children with ASD exhibit significant delay and impairment in
language and reading. The presence of language impairment in children with ASD exacerbates social
impairment and widens the achievement gaps in school. Despite the urgency of identifying and treating language
impairment in children with ASD, the critical gap in our understanding of the origin of the language variability in
ASD remain as the major challenge. Statistical learning (SL), the robust human ability to implicitly learn and
adapt to regularities from inputs, has gained increasing attention in the field of atypical language development.
Given deficits in social interaction in ASD, it has been postulated that language acquisition in this population
capitalizes on implicit learning, such as SL, rather than explicit learning. However, due to the substantial
methodological challenges in both SL measures and research with a heterogeneous population, there is a dearth
of longitudinal datasets in the field to determine the mechanistic role and clinical value of SL in language
development in ASD. Our central hypothesis is that the bidirectional relationship between SL and language
undergoes a mutual bootstrapping process, effectively a virtuous cycle. Under this framework, we predict that
weakness in SL underlies the exacerbation of language and literacy delay in a major subgroup of school-aged
children with ASD. Aim 1 proposes to specify the longitudinal relationship between SL and language/literacy
development in children with ASD from first grade (6 years old) to third grade (8 years old). Aim 2 will focus on
determining the longitudinal relationships between neural bases of SL and developing language networks in the
brains of children with ASD. Aim 3 will test whether linguistic SL measured using artificial languages is a proxy
for children’s sensitivity to real-world statistics using corpus data. The proposed study will yield critical knowledge
for developing diagnostic tools to characterize implicit learning ability in young children with ASD. The multimodal
longitudinal investigation, incorporating novel and theoretically motivated measures of SL and language
functions, will illuminate the cascading effect of abnormal learning on language and literacy development. The
findings will pave the way for future research that tests the therapeutic potential of implicit learning paradigms
for language intervention in a naturalistic setting.
