Abstract
Autism spectrum disorder (ASD), like many other neurodevelopmental conditions, arises from complex interactions between genetics and environmental factors. In the last 50 years, significant efforts have been made to identify biomarkers and risk factors that can improve our understanding of autism etiology. However, high heterogeneity in causes, symptomology, and developmental trajectories has made this task challenging. One strategy to combat heterogeneity is to characterize individuals based on their brain phenotypes, commonly referred to as "neuro-subtyping". In this study, we analyzed electroencephalography (EEG) recordings from 144 infants aged 6-7 months and employed subtyping methods (latent profile analysis and hierarchical clustering) to identify subgroups. Our analyses revealed three distinct subgroups based on various language-associated EEG measures. We found that group membership was predictive of expressive and receptive language trajectories based on the Mullen Scales of Early Learning, with infants displaying high connectivity in language regions and lefthemisphere lateralization achieving the highest scores, while infants with overactivation of connectivity in the auditory network achieved lower scores. Notably, EEGderived subgroups did not predict a later ASD diagnosis, suggesting a lack of evidence for an ASD-specific phenotype during the first year of life. Results from this project contribute to a large body of research that supports using stratification approaches to decode heterogeneity in autism and its role in predicting behavioural outcomes.