Toddlers and Noisy Speech
Kids are awesome at picking up words. They snarf them up at an astonishing rate, like 10 a day for an average 2 and a half year old. But let’s say you want to work on how exactly they go about doing that. What would you do? How would you test your hypothesis? Maybe you’d have them come into your lab and run them through your experiment, or maybe you’d make the trip to their home, and work with them there, to see what words they learn more quickly.
But regardless of where you do it, the how you’re probably picturing in the way researchers in the past have: a quiet space in which you check how the kid is learning. And yet, that’s really not a good fit for the way in which kids actually learn. If everything had to be quiet, that’d be a big challenge for picking up words at the rate kids do. Generally, kids are having to fish words out of noisy speech streams - a TV in the background, other people talking, music, sounds from outside, lots of things. So how does the learning happen? What’s the threshold for picking out the words from the environment?
Well, there’s a cool new study from Brianna McMillan and Jenny Saffran that looks at how kids deal with more noise around in their word learning adventure. Specifically, they played recorded background noise at different decibel levels behind the words they were interested in whether the kids would learn. For the first part of the study, 40 22-24-month-old kids were played nonsense words like tursey a few times over the course of an otherwise normal paragraph, so they could get used to hearing it; then shown a picture while hearing the novel word as a label for it; and finally, getting shown two pictures, the one they’d trained on and a new one, and being directed to try to find the one corresponding to the word (Look at the tursey!). But most importantly, for the first two test sections for each word, the speech and the label was played over background noise - half heard speech that was a quiet 5 dB difference from the background noise, and half got a larger 10 dB difference, with louder speech.
The researchers found that in the 10 dB difference case, the kids were able to focus on the target picture when the relevant word played. So when the labeling speech stood out that much from the background, they picked it up. But when the label was only 5 dB different from the background, the toddlers didn’t learn anything. They didn’t pick up the relevant word, because it didn’t stand out enough.
Not only that, but the researchers decided to look at some older kids: maybe with an extra half-year on them, 28-30-month-olds could pick out the relevant words even at a 5 dB difference. After all, they know a lot more words, so maybe it’ll be easier at that point, since they’re good at learning words by then? Plus some previous research suggests that kids get better at picking familiar words out of background noise as they age, so maybe that would extend to word learning, too. But alas, it turned out that nope, 10 dB may still be enough of a difference, but 5 dB apart from the background noise doesn’t get kids anywhere, using the same tests as with the younger kids.
But then the researchers thought... what if kids had heard the words clearly before, only in fluent speech - no labels, not trying to attach to anything, they’d just heard it around? Would that help them out? There’s previous research suggesting that hearing new words in fluent speech does help later on, but no one had really checked before whether that kind of experience could overcome background noise.
So in a third experiment, they gave 26 English-speaking 27-30 month old toddlers the same tests as before, but with a new twist: now, in the first phase, where they heard the new words they were supposed to pick up, for half of the words, that section was presented without any background noise. And then, the researchers only presented the new words with the picture labels with the higher background noise that kept them from learning before. So the only point where kids could learn what a tursey actually was had so much noise, kids in the previous parts of the experiment couldn’t pick out what they were supposed to be learning.
And lo and behold, here, our toddlers managed to pick up the new words just fine... as long as they’d heard them clearly before. Even though the words’d been heard out of context, with nothing to connect that bunch of sounds to, as long as they had some quiet exposures to the words, then they were able to learn the words, even if they weren’t really that differentiated from background noise.
It’s important to note that the size of the study is pretty small - the number of novel words each child was exposed to was quite low, and it’d be good to get some more generalization on this in future work. It’d also be good to know how long between encountering the word and actually learning it we can find an effect for. But what this suggests for now is that kids can still learn new words, even with challenging levels of background noise, as long as they do get some quiet exposures at some point to hold on to. Which can definitely help in a time where there’s a lot of distractions. ^_^
