Date created: | Last Updated:

Category: Project

Description: A key design feature of human language is the unbounded composition of meaningful units, such as words, intonations, and gestures. The ontogenetic development of this multimodal capacity has remained elusive. Using machine learning, naturalistic recordings, and cross-cultural experiments we show that prelinguistic infants compose pointing gestures with specific types of vocalizations. These compositions successfully differentiate statements from requests. The intended meanings are recognized not only by caregivers and bystanders from the same linguistic and cultural background as the infants, but also by perceivers who do not share any commonality with them. This suggests that prelinguistic compositions of gesture and vocalizations are a universal foundation on which full-blown compositionality can develop in humans.