Periodic Reporting for period 2 - ExELang (Experience Effects on early language acquisition)
Reporting period: 2023-05-01 to 2024-10-31
The project is focused on understanding how young children learn language in different parts of the world by recording their daily speech and what they hear. This is important because early language skills play a key role in children's development and future success, but most research has been done in wealthy, Western countries. Our goal is to collect and analyze data from children in a wide range of environments to learn how they acquire language, even in places with fewer resources. At the heart of our work are recording devices that capture audio over many hours, allowing us a child-centered view of what children in different places say and hear in a typical day. By using this kind of very naturalistic data, we hope to allow theory and practice to be based on more solid, more diverse evidence. We believe such an evidence base is necessary for our societies to develop tools for education and policy that can support all children in learning to speak and communicate effectively.
Since the start of the project, we’ve made significant progress in understanding how children learn language in diverse environments. We’ve collected a large amount of data from over 1,000 children in 14 different countries, using special devices that record the speech children hear and produce throughout the day. This data is being used to study the spoken input children receive and how it influences their language development.
One key finding so far is that children can learn multiple languages from less speech input than previously thought. For example, children in a remote community in Vanuatu hear less speech than children in the U.S. yet they are exposed to more languages and still show similar language development patterns. This challenges the common belief that more speech always leads to better language skills. We’ve also worked on developing new tools that help us analyze this data more efficiently, including machine learning methods that allow us to handle large amounts of recordings.
One key finding so far is that children can learn multiple languages from less speech input than previously thought. For example, children in a remote community in Vanuatu hear less speech than children in the U.S. yet they are exposed to more languages and still show similar language development patterns. This challenges the common belief that more speech always leads to better language skills. We’ve also worked on developing new tools that help us analyze this data more efficiently, including machine learning methods that allow us to handle large amounts of recordings.
Our project has pushed the boundaries of what we know about early language development by using long-form audio recordings to gather data from children in many different countries and cultures. This approach allows us to study language learning in real-world settings, rather than controlled labs, and with much larger datasets than ever before. We’ve also developed new technologies, such as machine learning tools, that make it possible to analyze these massive recordings, which would otherwise take years to process manually. By the end of the project, we expect to have a deeper understanding of how children around the world learn language, as well as improved tools for analyzing speech data. These tools could help researchers, educators, and policymakers better support children's development, particularly in under-resourced communities.