<b>A Data-Driven Exploration of Phonological and Morphological Processing in English Monolingual, Chinese-English Bilingual, and Spanish-English Bilingual Children Using Unsupervised Machine Learning (Silhouette = 0.5546)</b>

This paper presents an unsupervised machine learning analysis of a large-scale dataset examining morphological and phonological processing in English monolingual, Chinese-English bilingual, and Spanish-English bilingual children. The dataset is drawn from a functional Near-Infrared Spectroscopy (fNIRS) and behavioral study of 343 children (plus additional entries for missing or complementary records), covering comprehensive linguistic measures (phonological awareness, morphological awareness, vocabulary, word reading, nonword reading, passage comprehension, and reading fluency), as well as demographic and language-background data.Leveraging four primary clustering methods—K-Means, Agglomerative Clustering, DBSCAN, and HDBSCAN—we identified distinct grouping structures within the data based on multiple numeric features. A key objective was to investigate whether unsupervised approaches could meaningfully differentiate children based on their linguistic or literacy profiles, potentially uncovering subgroups pertinent to reading development. The best overall silhouette score (0.5546) was obtained by Agglomerative Clustering with two clusters using ‘average’ linkage.Notably, we contextualize these findings within the broader literature on bilingual language acquisition and reading-related cognitive processes in childhood. Our results encourage a deeper examination of emergent language and reading constructs using advanced unsupervised methods and underscore the dataset’s utility for exploring bilingualism, morphological and phonological skills, and reading development. <br>