Parents and Children Jointly Create Multimodal Semantic Regularities During Naturalistic Toy Play.

Early language learning involves not only learning the meanings of individual words but also learning semantic connections among words. Prior research has documented children's sensitivity to different semantic relations, yet little is known about how this semantic knowledge is acquired from everyday interactions. This study examined the semantic regularities children and parents jointly created during unstructured, naturalistic toy play. Forty-four parent-child dyads engaged in a free-play session with 27 toys spanning three categories (food, animals, vehicles). Head-mounted eye-trackers recorded children's eye gaze transitions to determine how children and their parents jointly selected which toys to play with. Parents' concurrent speech was also extracted. Results revealed that children were more likely to select a next toy to play with from the same category as the current toy than from other semantic categories, suggesting that toy play is characterized by temporal sequences of semantically-related toys. Additionally, children heard more unique words from the same semantic category co-occur during these same semantic category toy transitions. These findings highlight that semantic structure is not merely present in parents' speech-it is constructed by parents and children through shared attention and action, advancing our understanding of how children use their limited input statistics to build rich semantic knowledge.