Mapping the Communicative Landscape of Early Child-Caregiver Dialogue

Early linguistic interaction plays a key role in children's social and cognitive development. However, there is a lack of quantitative studies that offer a comprehensive insight into the communicative landscape that characterizes child-caregiver dialogues.In this study, we apply advanced Natural Language Processing (NLP) techniques to analyze multiple corpora, encompassing data from N = 609 individual children (aged 20 to 32 months old), over 2,500 conversations and around 700k pairs of child-caregiver interactive turns. Our methodology combined models for communicative intent inference and models for semantic coherence evaluation, providing a comprehensive analysis of linguistic coordination and its emergence in infancy. We found that both focused turn pairs with clear communicative intents (e.g., question-response) and open turn pairs (e.g., statement-statement) were common in structuring interactions. However, these pairs varied significantly regarding their semantic coherence (e.g., open pairs were generally less coherent) and role (e.g, caregivers initiating most of the focused pairs). Additionally, we document developmental shifts in intent expression and, interestingly, a dissociation between frequency vs. coherence. The broad impact of this study lies in establishing a solid, bottom-up empirical foundation for developing ecological theories of communicative development.