Maturity of Baby Sounds

The Maturity of Baby Sounds

---

Did you know that, well before saying real words, children produce very different sounds with their mouths? Already at birth, they cry and cry. We all know these sounds! A little later, they also begin to babble - that is, produce sounds closer to those of adult speech, because they have nice and clear consonants (like "t" and "p") and vowels (like "a" and "i"). These are the ones that interest us the most.

We know that these speech-like sounds become more prevalent as babies grow older, but not a lot more. For instance, we don't know whether this happens at similar rates regardless of the language the baby is learning, nor when they begin to match adults in terms of how frequently they make them. We are also interested in seeing whether sounds babies make are different in infants that go on to have a medical diagnosis, such as Autism.

Today, most of the babbling data for babies comes from the United States and from typically-developing babies. With your help, we will be able to generate data for many more languages, cultures, and populations. Here is a world map with some of the locations in which our data were gathered:

Our data are really special because the child him/herself wears a recorder all day long. Then we analyze these 10-16h long recordings with software to try to find when the child or someone else is making sounds with their mouth. We cut our clips really short to protect the identity and privacy of the participating families.

Also, we find where there may be vocal sounds using a piece of software. When the software is wrong, you may hear an animal or just noise. By tagging that these are errors, you are also helping us improve our automated software. In fact, our software sometimes confuses female adults and babies! So your tags will also help us distinguish babies from other children, female adults, and male adults.

As volunteers like you label these clips, we can look at what proportion of the clips contain speech sounds rather than purely emotional sounds (crying and laughing). And we can also look at what proportion of those speech sounds are "complex" (what we call canonical in the field guide) -- in speakers of different ages.

We launched this project in 2020 with an initial workflow classifying baby sounds. This let us ask questions like:

• Do babies everywhere produce more speech and fewer crying sounds as they grow?
• Is the ratio of more versus less complex speech sounds the same regardless of which language the baby is learning? Or are there differences between children learning English versus, for instance, a Pidgin language that has only 10 consonants, or a Khoisan language that has 100 consonants?
• Do these changes happen at the same rate in children at risk of language impairment?

Because of your help, we’ve found that typically-developing children’s speech development is very similar across languages and cultures! We've also found that children at risk of language impairment show different patterns, which could help in identifying intervention points. So, a big thank you!

We’re now relaunching the project with a new workflow that asks you to not only classify baby sounds, but also the sounds of the people around them. This will let us better study children’s speech development and how it relates to their caregiver’s speech. With this new workflow, we are looking forward to answering these questions:

• When do children match adults in the frequency with which they make complex sounds? Does this depend on the language they are learning? How does this differ between typically developing children and children at risk for language impairment?
• Does children’s speech development depend on how much speech and what kind of speech they hear from the people around them?
• Are some of our recorders and/or software better than others? How well do they handle child speech vs. adult speech?