How Babies Learn Words

These are rough times.  If you’re like me, you’re spending too much time reading the news and fretting about the uncertainty of the months ahead.  That, and trying to keep the kids from seriously wounding each other.  Kind of like our collective reaction to the coronavirus, my kids’ disputes tend to escalate rather quickly.

So, what to do at times like this?  For starters, we’re following the directions of health officials and parking ourselves at home.  This time feels truly momentous; it’s a time that begs for reflection.  Time to take stock of where we are and where we want to be in our lives, whenever this is all over.  And while I’m not entirely sure where I’m going, I do know where I’ve been.  So, I want to share some things about language development that are so cool that when I first learned about them, I knew I was heading in the right direction. 

I first learned about statistical learning many years ago in a very, very large undergraduate psychology class at McGill.  If you’ve taken this type of class, you know that they have the potential to be a bit unengaging due to their size, yet Dr. Athena Vouloumanos, who has published lots on this topic, was totally inspiring. She just seemed so passionate about child language development, and the day she talked about statistical learning, I was hooked. 

Statistical learning is when we notice regularities in the world around us.  Humans have a remarkable ability to go around taking statistics on things they observe.  For example, we observe that most women (around 95% of us) are between 4’11” and 5’10” tall and so a woman who is 6’0” seems quite tall.  Noticing patterns and regularities is what helps babies make sense of things after they come crashing into a world full of unfamiliar experiences. It turns out that babies are especially good at taking stats on the language(s) they hear around them.  Sentences are actually spoken in a stream of connected sounds, and so when you say “oh what a handsome baby you are”, the baby actually hears “ohwhatahandsomebabyyouare” (AND the sounds even overlap with each other, in a phenomenon known as coarticulation).  If it’s not obvious where one word ends and another begins, how does a baby figure out what words are?

Incredibly, to figure this out, babies use their statistical observations about which sounds are more likely to occur in succession.  The earliest study of this was by Saffran, Aslin, and Newport in 1996.  They used the term “transitional probability” to refer to the likelihood that one particular syllable will be followed by another particular syllable.  For example, babies often hear the word baby as in “what a sweet baby” and “look at the little baby toes.”  They observe that the spoken syllable “bay” is very often followed by the syllable “bee, and so is likely a whole word (baby).  What follows “bee” is highly variable, and therefore must be the start of a different word. 

The researchers tested this by having 8-month-olds listen to a 2-minute string of synthesized speech: four different 3-syllable nonsense words, bidaku, padoti, tupiro, golabu, presented in random order in steady, monotone syllables.  Every time the baby heard bi, it was followed by da (because they’re in the same word), but when the baby heard ku (the last syllable of bidaku), it was followed by one of three different syllables, marking the start of a new word.  That is, the transitional probability of bi-da was 100%, whereas the transitional probability of ku-pa was only 33%.  Then, after the two minutes of listening was over, the babies were presented with either a familiar word (that was part of the 2-minute listening, e.g. bidaku), or a novel word made by combining syllables from different words, e.g. bitula.  Unbelievably, the babies were able to tell the difference between the two kinds of words.  At 8 months old!  After 2 minutes of learning!  Incredible.  The authors also found similar results with adults, though they listened for 21 minutes in order to learn the transitional probabilities of the syllables. 

my statistical learner

As someone who learned a second language as a young adult, I have some experience with this phenomenon in real life, except it took me longer than 2 (or 21) minutes.  I was fortunate to learn French in a social setting, where I was often the only anglophone in a group of Québécois who talked a million miles an hour.  There are several phrases for which I had a general sense of the meaning, but no idea what the individual words were.  These include: aunmomadoné, antuca, and bewayondon, which, after a long time, I was able to parse into à un moment donné (at some point in time), en tout cas (in any case), and bien voyons donc (holy cow).  I’m still not totally clear on how exactly feck breaks down into ça fait que (and so…), but that’s just one of the charms of Québécois French.

Back to the babies for a moment.  Were you wondering how we know that babies know the difference between novel and familiar words?  Because babies can’t tell us what they observe, researchers need to tap into behaviours that they can do.  Babies are good at two things: looking at things, and sucking on things.  Researchers use a procedure called habituation where they expose the baby to repetitive stimuli (like 2 minutes of monotone syllables), until they get a little bored.  Then, they present the “test” stimuli and observe the baby’s reaction.  When the baby notices something new, he/she will perk up and suck quicker on a pacifier, or look longer toward the stimuli.  Of course you can’t really conclude much from one baby’s response, so tests are repeated with many babies.  Saffran, Aslin, and Newport found that among their group of 24 babies, the ones who were presented with novel words looked significantly longer, indicating they had learned the words during the 2-minute habituation.  In conclusion, babies are geniuses.

Now, is this useful knowledge? Absolutely. When talking with babies, it’s a great idea to say new words many times in different phrases, and even emphasize the new word a little bit to help it stand out. If you’re playing with an airplane, you might say “What a big airplane!” and “The airplane flies so fast!” and “Do you like my blue airplane?” This repetition, combined with lots of opportunities to observe how words are paired with whatever they represent (like “airplane” with toy airplanes, real airplanes, airplane pictures) is basically how babies learn words. Before babies can say their first words, typically around 12 months of age, they have already learned to understand a lot of words in this way.

I hope this has been an interesting distraction in this very difficult time, especially if you find yourself at home with little language learner.  Hang in there, friends, and if anyone wants to chat about any topics mentioned on this blog, don’t hesitate to reach out. Goodness knows, we can use the distraction.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: