On learning the sounds of a new (spoken) language

By Grant Jacobs 17/02/2011

Infants learn languages so easily. In particular, they are able to distinguish the important sounds in a language.

Below are two videos of lectures given by Professor Patricia Kuhl (University of Washington) exploring how infants learn language.

The first first, the shorter of the two, is a TEDx presentation; the latter is a longer exposition given at the University of Washington. Both are aimed at a general audience.

You may wish to skip the introductions in the second talk: start six and half minutes in if you’d rather just see ‘the meat’. This second talk is a little older: it’s from two years ago, but it also has more detail. (And wordier, but that comes with the setting she is speaking in.)

One word about her TEDx talk before watching. In it Prof. Kuhl shows a slide of the drop-off language learning ability as we age. As she explains more fully in her longer talk (at about 15 minutes in) this presents the ability to learn new language sounds from a second language. This is not quite the same as learning to speak a new language as an adult. One of the commenters, Yuhfen Lin, put it better than I could:

What she means is that as a person grow old, the ability [to] differentiate sounds in [a] different language decrease dramatically after puberty. It is possible to learn pronunciation (how to pronounce a new sound) if you have a good language teacher. But it is very hard to learn to hear the sound even when you know how to produce it.


I guess you’re wondering why I emphasised ‘spoken’ in my title.

There are sign languages, too: visual languages as used by many deaf communities.*

Infants learn these ’natively’ as well, processing visual signals rather than auditory ones.

Sounds are tricky things for an infant to make accurately: simple ’proto-signs’ are easier. An upshot is that quite young infants are able to make simple communications with adults using simple signs.

I’ve experienced this at first-hand. Even with a little sign language experience (as I do), it’s surprising the first time you see it.

You’ll have all heard of ’baby sign’, promoted as a way to communicate with young children. This popular video, with over 3 million views, gives you an idea of what I mean:

Professor Kuhl talks about software research projects aimed a spoken interfaces for computers in her longer presentation. There’s also work on sign language interfaces, too. (I have previously shown a video of direct brain interfaces: do watch that if you haven’t already – the future is looking remarkable, you’d have to say.)


I’m not a specialist in this area, but it’s fascinating.

* Aside from a casual interest in sign languages, another connection for me is a talk I gave as part of a fourth-year psychology course I sat in on a number of years ago that looked at early studies of brain activity during processing of signed languages and the background reading I did for this.

Other articles on Code for Life:

Consumer brain-computer interface

Understanding the brain through controlling it

Backstage at the Diamond Light Source

Temperature-induced hearing loss

Autism — looking for parent-of-origin effects

0 Responses to “On learning the sounds of a new (spoken) language”

  • My 1 yr old is growing up bilingual (English and German), and I often wonder if learning two very similar languages is more difficult to sort out than two very different ones (like say English and Mandarin). I wonder if when the sounds and syntax of the languages are similar, does she even realise its 2 different languages? I’m sure she will eventually, but maybe it will take a bit longer – time will tell when she starts to speak I guess.

    That sign language thing is quite inspirational – but I suspect requires a large time investment on the part of the parent! I suspect I would be too busy/distracted to be consistent enough for her to pick it up!

  • Good question. I don’t know the answer!

    About SLs, it would require effort on your part – because you don’t know SL. I doubt it’s any more effort for the kid! Perhaps one way to thing about it is to consider hearing children of signing deaf parents. They grow up natively bilingual in sign and the local spoken language. Have to rush off so no time to elaborate right now.

  • Erratum: ‘think’ for ‘thing’, sorry!

    One interesting sideline is that SLs (sign languages) are “true” languages in the sense that they are processed by the brain in the language areas. I gave a talk on this, presenting the first papers showing this, as part of a 4th year psychology paper I sat in on years ago. I’d have to check to see what’s happened in this area since (so I don’t say something terribly wrong), but if I could track down my old talk material it might make for an interesting blog post (if a little dated from a science point of view).

    It’ll be interesting to see how your kid gets on with both languages. Is she (he?) getting both languages from both parents? Anecdote: I knew a kid whose mother spoke English and French, and whose father spoke English and Hebrew. An interesting side effect was the little one initially would speak French to women and Hebrew to men, but rarely vice versa. It would be interesting to check if this is common in trilingual families. (It was amusing at times. The little one would often start off in Hebrew to me, which I don’t know, and I’d have to coax the kid into speaking English. Little smarty-pants.)

  • You are right, Grant, that the longer video has more substance to it. Thanks for bringing it to my attention. I still object to the term ‘taking statistics’ however. Language has statistical properties, yes. But there is nothing in her presentation that shows that infants learning language are engaged in anything more complicated or abstract than stimulus-response. Perhaps she does not mean to imply anything more than that with the phrase, but at best then it’s an empty phrase and at worst she is implying more than her research shows. It is the kind of phrase meant to excite interest without actually conveying information. PR not science. Perhaps it’s what one has to do to get grant money, but I have a strong adverse reaction to it that I can’t just shrug off. Sorry if that makes me a killjoy.

  • Hi Jamie,

    Her reference to “taking statistics” will be to other work. There is a sizeable literature of research examining language.

    The experiments she shows in her talks of the kids responding to sounds don’t test the learning process itself, but what the kids have learnt earlier. The learning process, the ‘taking statistics’ that she refers to, happens during their daily lives before those tests. Because they are testing what has been learnt previously, it will be stimulus-response. They’re asking “do they recognise these language elements?” – asking if they have learnt them during being exposed to language.

    (There is subtlety in ensuring that the experimental design avoids “accidentally” testing only for short-term term learning during the test itself.)

    The phrase ‘taking statistics’ she uses will refer to studies showing that infants are in effect language motif detectors! (I’m crossing disciplines here: the term ‘motif’ is really from my own field, computational biology. Use vowels or phonemes if you prefer.)

    As I briefly mentioned on Greg Laden’s blog, imagine a stream of sound, someone speaking rapidly. I find it easiest to think of the stream as the squiggly line you see in a graph of a sound recording. The word stream is important – think of a continuous stream of complex sounds, no breaks between what we, as adults, have learnt are ‘language elements’.

    A brain learning language identifies the repeated motifs within the stream. You can think of it as a sort of frequency analysis. What short stretch of sound (say, a vowel) tends to ‘hang together’ as a motif, with differing sounds before and after them. They’re picking out the basic elements of the language from the stream of sound (without knowing that language (yet) or the meanings of the words).

    It’s made more complicated by that different people speaking the same language speak differently, as she referred to.

    Picking out these smaller elements of language from the stream of sound is more that stimulus-response. They’re not getting those in nice tidy ‘chunks’ they can relate to something, but elements within a stream of sound. Furthermore, at that age they don’t yet have meaning pinned to the words, etc. They’re just “bits of sound” (motifs, vowels, etc.) that their brain recognises as being part of this language it is learning.

    Do you mind asking me what your background is? I’ve assumed you’re new to this. (Excuse me if you’re not.)