From Formulaic Chunks to Productive Syntax: A Cross-Linguistic Network Analysis

Children transition from formulaic, item-based chunks to flexible, productive syntax. To quantify this structural shift, we modeled 3.9 million utterances from 14 mother-child dyads across five languages (German, English, Hebrew, Dutch, and Swedish) as bigram-based networks. Using global network transitivity to measure "cliquiness," we tracked the dissolution of formulaic islands. Results revealed a robust cross-linguistic pattern: network transitivity significantly declined as vocabulary grew, marking the structural onset of productivity. Crucially, we found a U-shaped relationship between utterance length and connectivity. High formulaic usage persisted in very short and very long utterances, suggesting children revert to holistic chunks to bypass processing bottlenecks when producing complex speech. We also found children's networks are finely tuned to their input's structural properties. These findings support usage-based accounts, offering a precise quantitative signature for the emergence of productivity.