Infants' learning from the interactive speech environment

Many language problems that arise in the course of child development can be traced back to children’s early ability to process and make sense of speech. This project studies how infants begin learning their language by using new methods to evaluate infants’ first steps in understanding the speech of their parents. This project combines observational methods and computer modeling to discover which aspects of speech to children are most important in helping them to learn language. This research will result in creation of a large, annotated database of child directed speech that can be shared broadly with other researchers to accelerate further discoveries about infant language development. This research project consists of two related studies. In the first study, parents and infants play together with toys while their speech is recorded by researchers. Infants wear a head-mounted eye tracker that records where they are looking. Infants tend to look at things people talk about, therefore their eye movements make it possible for researchers to know when speech is interpretable to infants and when it is not. In the second study, investigators create and annotate a large corpus of everyday spoken language. Researchers then use state-of-the-art speech recognition technology to estimate what children learn from the speech they hear at different ages. These estimates are then evaluated by examining which words children have learned. This project uses realistic interactions between children and caregivers, and computer modeling tools to advance understanding of early language development. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.