A new wearable audio recorder with multiple microphones will capture the precise sounds babies hear at home, letting researchers measure whether exaggerated vowel sounds in “baby talk” actually boost language learning. Most studies of early language rely on recordings too noisy to analyse the fine acoustic details of speech—the stretched vowels, the clear consonants, the slowed tempo that adults instinctively use with infants. This project builds a better recorder, then tests whether those phonetic features predict vocabulary growth in children from multilingual homes, starting at nine months old. The same device will track adults learning English after moving to the UK, asking whether the same acoustic patterns help older learners too. If the recorder works as designed, it will give language researchers a cheap, open-source tool to study speech input at a level of detail previously impossible outside a laboratory. The immediate payoff is fundamental: a clearer picture of how the sound of speech—not just the number of words—shapes learning across the lifespan. That understanding could eventually inform how parents, teachers, and language apps structure spoken input for children and adult learners alike.
View original technical description
Children who hear more socially engaging speech early in life have been found to have larger vocabularies later in childhood, which is a discovery that may help improve language and educational outcomes in children. The central hypothesis of our research is that input can also help explain learning throughout the lifespan (e.g., why adults have more difficulty than children learning new languages) and we can therefore gain a fuller view of speech learning by extending the methods used in child speech input research to older ages. To accomplish this work, we are developing a new wearable audio recorder that is designed to better measure speech at a phonetic level. Child speech input research typically uses wearable audio recorders to quantify input, but the recorders tend to produce recordings that are too noisy for the phonetic analysis of speech. In our project, we will develop a new open-source wearable recorder that uses microphone arrays to better isolate speakers from background noise, allowing us to measure how the acoustic characteristics of speech input (i.e., phonetic aspects, such as exaggerated differences between vowels) lead to greater learning. We will conduct longitudinal investigations with children from multilingual homes starting at 9 months old, and similarly test multilingual adults who move to the UK for the purpose of education. The acoustics and words of speech input will be measured using our recorders and analysed using corpus phonetics techniques, machine learning, and AI. Speech perception skills will be assessed using EEG analyses of auditory and lexical processing, using naturalistic paradigms in which the subjects listen to ordinary podcasts. We will also include assessments of attitudes towards languages and accents, as a counterpoint to our focus on speech input and development. Our project will thus add to the technologies and methods available to speech input researchers, extend the investigation to understand speech learning throughout the lifespan, and produce a more comprehensive view of language learning at the phonetic level.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know