This project aims to discover how infants begin to learn words and speech sounds in the first year. When infants first hear speech, they do not know what parts of each sentence correspond to individual words. They probably do not even have the idea that words exist, but must discover it. This project uses what is known about infants' speech perception and memory to test what infants can learn from the speech signal. This will be done by making and annotating recordings of mothers at home speaking to their infants. The researchers will carefully measure the acoustics of words and speech sounds like consonants and vowels, and will use computer models to estimate what language information could be available to the naive infant. By linking together these estimates, and infants' language accomplishments (like vocabulary) as toddlers, it is possible to test how the maternal language environment provides for infant learning, and will permit discovery of what features make learning easiest for babies. The project will measure a diverse sample of mothers and children from Philadelphia and from elsewhere in the country. Understanding how language learning begins in young infants is important for understanding variability in young children's language related outcomes. It is also important for giving parents guidance about how to encourage language skill even in their very young children. Ultimately this will give us a better understanding of the beginnings of language development.
This project aims to discover how infants begin to learn words and speech sounds in the first year. In doing so the project will flesh out and challenge the proposal that infants learn speech sounds by learning the forms of words, contra the dominant view that infant phonetic learning is initially independent of the nascent lexicon. Central to the project is the de novo creation of a longitudinal corpus of infant-directed speech in African-American (AA) families, and the coordinated annotation of lexical and phonetic features in both this AA corpus and in the mainstream-English Seedlings corpus. The project will allow characterization, in unprecedented detail, of the consequences of specific maternal behaviors for specific learning outcomes in children, down to the individual word level. The research will evaluate the degree to which these relationships hold in both low and high SES families. The project will also use sensitive laboratory methods, including eye-tracking analysis of infants' fixation to named pictures, to evaluate the phonetic specificity of infants' early representations of words. The result will be a quantitatively specified characterization of how infants begin to learn how their language works.
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.
