A dynamic network approach to bilingual child data

Abstract Usage-based approaches to language acquisition emphasize the central role of input–output relationships in the gradual emergence of linguistic knowledge. Recent computational work has provided empirical support for this view by showing how network-based analyses can capture constructional patterns in child language. One such method is the Dynamic Network Model (DNM). However, it remains unclear whether these findings extend to bilingual acquisition, where differing input conditions across the two languages are expected to shape the formation of linguistic networks more visibly. To address this gap, the present study investigates two German–English bilingual children aged 2;03–3;11. Applying the DNM to both child-directed speech and children’s own utterances, we examine how constructional “pivots” emerge under bilingual input conditions and how they differ between individual children. Our results indicate that the DNM identifies informative clusters in the data, reflecting the language distribution in the input and thus provides a promising heuristic for advancing our understanding of bilingual first language acquisition.