TEDxRainier
Patricia Kuhl: The linguistic genius of babies
Filmed:
Readability: 4.2
3,424,262 views
Patricia Kuhl shares astonishing findings about how babies learn one language over another -- by listening to the humans around them and "taking statistics" on the sounds they need to know. Clever lab experiments (and brain scans) show how 6-month-old babies use sophisticated reasoning to understand their world.
Patricia Kuhl - Language
Patricia Kuhl studies how we learn language as babies, looking at the ways our brains form around language acquisition. Full bio
Patricia Kuhl studies how we learn language as babies, looking at the ways our brains form around language acquisition. Full bio
Double-click the English transcript below to play the video.
00:15
I want you to take a look at this baby.
0
0
3000
00:18
What you're drawn to are her eyes
1
3000
3000
00:21
and the skin you love to touch.
2
6000
3000
00:24
But today I'm going to talk to you about something you can't see --
3
9000
3000
00:27
what's going on up in that little brain of hers.
4
12000
3000
00:31
The modern tools of neuroscience
5
16000
2000
00:33
are demonstrating to us that what's going on up there
6
18000
3000
00:36
is nothing short of rocket science.
7
21000
3000
00:39
And what we're learning
8
24000
2000
00:41
is going to shed some light
9
26000
2000
00:43
on what the romantic writers and poets
10
28000
3000
00:46
described as the "celestial openness"
11
31000
3000
00:49
of the child's mind.
12
34000
2000
00:51
What we see here
13
36000
2000
00:53
is a mother in India,
14
38000
2000
00:55
and she's speaking Koro,
15
40000
2000
00:57
which is a newly discovered language.
16
42000
2000
00:59
And she's talking to her baby.
17
44000
2000
01:01
What this mother --
18
46000
2000
01:03
and the 800 people who speak Koro in the world --
19
48000
3000
01:06
understands [is] that, to preserve this language,
20
51000
3000
01:09
they need to speak it to the babies.
21
54000
3000
01:12
And therein lies a critical puzzle.
22
57000
3000
01:15
Why is it that you can't preserve a language
23
60000
2000
01:17
by speaking to you and I, to the adults?
24
62000
3000
01:20
Well, it's got to do with your brain.
25
65000
3000
01:23
What we see here
26
68000
2000
01:25
is that language has a critical period for learning.
27
70000
3000
01:28
The way to read this slide is to look at your age on the horizontal axis.
28
73000
3000
01:31
(Laughter)
29
76000
3000
01:34
And you'll see on the vertical
30
79000
2000
01:36
your skill at acquiring a second language.
31
81000
2000
01:38
Babies and children are geniuses
32
83000
2000
01:40
until they turn seven,
33
85000
2000
01:42
and then there's a systematic decline.
34
87000
3000
01:45
After puberty, we fall off the map.
35
90000
2000
01:47
No scientists dispute this curve,
36
92000
3000
01:50
but laboratories all over the world
37
95000
2000
01:52
are trying to figure out why it works this way.
38
97000
3000
01:55
Work in my lab is focused
39
100000
2000
01:57
on the first critical period in development --
40
102000
2000
01:59
and that is the period in which
41
104000
2000
02:01
babies try to master which sounds are used in their language.
42
106000
3000
02:04
We think, by studying how the sounds are learned,
43
109000
3000
02:07
we'll have a model for the rest of language,
44
112000
2000
02:09
and perhaps for critical periods that may exist in childhood
45
114000
3000
02:12
for social, emotional
46
117000
2000
02:14
and cognitive development.
47
119000
2000
02:16
So we've been studying the babies
48
121000
2000
02:18
using a technique that we're using all over the world
49
123000
2000
02:20
and the sounds of all languages.
50
125000
2000
02:22
The baby sits on a parent's lap,
51
127000
2000
02:24
and we train them to turn their heads when a sound changes --
52
129000
2000
02:26
like from "ah" to "ee."
53
131000
2000
02:28
If they do so at the appropriate time,
54
133000
2000
02:30
the black box lights up
55
135000
2000
02:32
and a panda bear pounds a drum.
56
137000
2000
02:34
A six-monther adores the task.
57
139000
2000
02:36
What have we learned?
58
141000
2000
02:38
Well, babies all over the world
59
143000
2000
02:40
are what I like to describe
60
145000
2000
02:42
as "citizens of the world."
61
147000
2000
02:44
They can discriminate all the sounds of all languages,
62
149000
3000
02:47
no matter what country we're testing and what language we're using,
63
152000
3000
02:50
and that's remarkable because you and I can't do that.
64
155000
3000
02:53
We're culture-bound listeners.
65
158000
2000
02:55
We can discriminate the sounds of our own language,
66
160000
2000
02:57
but not those of foreign languages.
67
162000
2000
02:59
So the question arises:
68
164000
2000
03:01
when do those citizens of the world
69
166000
2000
03:03
turn into the language-bound listeners that we are?
70
168000
3000
03:06
And the answer: before their first birthdays.
71
171000
3000
03:09
What you see here is performance on that head-turn task
72
174000
3000
03:12
for babies tested in Tokyo and the United States,
73
177000
2000
03:14
here in Seattle,
74
179000
2000
03:16
as they listened to "ra" and "la" --
75
181000
2000
03:18
sounds important to English, but not to Japanese.
76
183000
3000
03:21
So at six to eight months the babies are totally equivalent.
77
186000
3000
03:24
Two months later something incredible occurs.
78
189000
3000
03:27
The babies in the United States are getting a lot better,
79
192000
2000
03:29
babies in Japan are getting a lot worse,
80
194000
2000
03:31
but both of those groups of babies
81
196000
2000
03:33
are preparing for exactly the language that they are going to learn.
82
198000
3000
03:36
So the question is: what's happening
83
201000
3000
03:39
during this critical two-month period?
84
204000
2000
03:41
This is the critical period for sound development,
85
206000
2000
03:43
but what's going on up there?
86
208000
2000
03:45
So there are two things going on.
87
210000
2000
03:47
The first is that the babies are listening intently to us,
88
212000
3000
03:50
and they're taking statistics as they listen to us talk --
89
215000
3000
03:53
they're taking statistics.
90
218000
2000
03:55
So listen to two mothers speaking motherese --
91
220000
3000
03:58
the universal language we use when we talk to kids --
92
223000
3000
04:01
first in English and then in Japanese.
93
226000
3000
04:04
(Video) English Mother: Ah, I love your big blue eyes --
94
229000
3000
04:07
so pretty and nice.
95
232000
3000
04:11
Japanese Mother: [Japanese]
96
236000
6000
04:17
Patricia Kuhl: During the production of speech,
97
242000
2000
04:19
when babies listen,
98
244000
2000
04:21
what they're doing is taking statistics
99
246000
2000
04:23
on the language that they hear.
100
248000
3000
04:26
And those distributions grow.
101
251000
3000
04:29
And what we've learned
102
254000
2000
04:31
is that babies are sensitive to the statistics,
103
256000
3000
04:34
and the statistics of Japanese and English are very, very different.
104
259000
3000
04:37
English has a lot of Rs and Ls.
105
262000
3000
04:40
The distribution shows.
106
265000
2000
04:42
And the distribution of Japanese is totally different,
107
267000
2000
04:44
where we see a group of intermediate sounds,
108
269000
3000
04:47
which is known as the Japanese "R."
109
272000
3000
04:50
So babies absorb
110
275000
2000
04:52
the statistics of the language
111
277000
2000
04:54
and it changes their brains;
112
279000
2000
04:56
it changes them from the citizens of the world
113
281000
2000
04:58
to the culture-bound listeners that we are.
114
283000
3000
05:01
But we as adults
115
286000
2000
05:03
are no longer absorbing those statistics.
116
288000
2000
05:05
We're governed by the representations in memory
117
290000
3000
05:08
that were formed early in development.
118
293000
3000
05:11
So what we're seeing here
119
296000
2000
05:13
is changing our models of what the critical period is about.
120
298000
3000
05:16
We're arguing from a mathematical standpoint
121
301000
3000
05:19
that the learning of language material may slow down
122
304000
3000
05:22
when our distributions stabilize.
123
307000
2000
05:24
It's raising lots of questions about bilingual people.
124
309000
3000
05:27
Bilinguals must keep two sets of statistics in mind at once
125
312000
4000
05:31
and flip between them, one after the other,
126
316000
3000
05:34
depending on who they're speaking to.
127
319000
2000
05:36
So we asked ourselves,
128
321000
2000
05:38
can the babies take statistics on a brand new language?
129
323000
3000
05:41
And we tested this by exposing American babies
130
326000
2000
05:43
who'd never heard a second language
131
328000
2000
05:45
to Mandarin for the first time during the critical period.
132
330000
3000
05:48
We knew that, when monolinguals were tested
133
333000
2000
05:50
in Taipei and Seattle on the Mandarin sounds,
134
335000
3000
05:53
they showed the same pattern.
135
338000
2000
05:55
Six to eight months, they're totally equivalent.
136
340000
2000
05:57
Two months later, something incredible happens.
137
342000
3000
06:00
But the Taiwanese babies are getting better, not the American babies.
138
345000
3000
06:03
What we did was expose American babies during this period
139
348000
3000
06:06
to Mandarin.
140
351000
2000
06:08
It was like having Mandarin relatives come and visit for a month
141
353000
3000
06:11
and move into your house
142
356000
2000
06:13
and talk to the babies for 12 sessions.
143
358000
2000
06:15
Here's what it looked like in the laboratory.
144
360000
2000
06:17
(Video) Mandarin Speaker: [Mandarin]
145
362000
22000
06:39
PK: So what have we done to their little brains?
146
384000
2000
06:41
(Laughter)
147
386000
3000
06:44
We had to run a control group
148
389000
2000
06:46
to make sure that just coming into the laboratory
149
391000
2000
06:48
didn't improve your Mandarin skills.
150
393000
2000
06:50
So a group of babies came in and listened to English.
151
395000
2000
06:52
And we can see from the graph
152
397000
2000
06:54
that exposure to English didn't improve their Mandarin.
153
399000
2000
06:56
But look at what happened to the babies
154
401000
2000
06:58
exposed to Mandarin for 12 sessions.
155
403000
2000
07:00
They were as good as the babies in Taiwan
156
405000
2000
07:02
who'd been listening for 10-and-a-half months.
157
407000
3000
07:05
What it demonstrated
158
410000
2000
07:07
is that babies take statistics on a new language.
159
412000
2000
07:09
Whatever you put in front of them, they'll take statistics on.
160
414000
4000
07:13
But we wondered what role
161
418000
2000
07:15
the human being played
162
420000
2000
07:17
in this learning exercise.
163
422000
2000
07:19
So we ran another group of babies
164
424000
2000
07:21
in which the kids got the same dosage, the same 12 sessions,
165
426000
3000
07:24
but over a television set
166
429000
2000
07:26
and another group of babies who had just audio exposure
167
431000
3000
07:29
and looked at a teddy bear on the screen.
168
434000
2000
07:31
What did we do to their brains?
169
436000
3000
07:34
What you see here is the audio result --
170
439000
3000
07:37
no learning whatsoever --
171
442000
2000
07:39
and the video result --
172
444000
3000
07:42
no learning whatsoever.
173
447000
2000
07:44
It takes a human being
174
449000
2000
07:46
for babies to take their statistics.
175
451000
2000
07:48
The social brain is controlling
176
453000
2000
07:50
when the babies are taking their statistics.
177
455000
2000
07:52
We want to get inside the brain
178
457000
2000
07:54
and see this thing happening
179
459000
2000
07:56
as babies are in front of televisions,
180
461000
2000
07:58
as opposed to in front of human beings.
181
463000
2000
08:00
Thankfully, we have a new machine,
182
465000
2000
08:02
magnetoencephalography,
183
467000
2000
08:04
that allows us to do this.
184
469000
2000
08:06
It looks like a hair dryer from Mars.
185
471000
2000
08:08
But it's completely safe,
186
473000
2000
08:10
completely non-invasive and silent.
187
475000
3000
08:13
We're looking at millimeter accuracy
188
478000
2000
08:15
with regard to spatial
189
480000
2000
08:17
and millisecond accuracy
190
482000
2000
08:19
using 306 SQUIDs --
191
484000
2000
08:21
these are Superconducting
192
486000
2000
08:23
QUantum Interference Devices --
193
488000
2000
08:25
to pick up the magnetic fields
194
490000
2000
08:27
that change as we do our thinking.
195
492000
2000
08:29
We're the first in the world
196
494000
2000
08:31
to record babies
197
496000
2000
08:33
in an MEG machine
198
498000
2000
08:35
while they are learning.
199
500000
2000
08:37
So this is little Emma.
200
502000
2000
08:39
She's a six-monther.
201
504000
2000
08:41
And she's listening to various languages
202
506000
2000
08:43
in the earphones that are in her ears.
203
508000
3000
08:46
You can see, she can move around.
204
511000
2000
08:48
We're tracking her head
205
513000
2000
08:50
with little pellets in a cap,
206
515000
2000
08:52
so she's free to move completely unconstrained.
207
517000
3000
08:55
It's a technical tour de force.
208
520000
2000
08:57
What are we seeing?
209
522000
2000
08:59
We're seeing the baby brain.
210
524000
2000
09:01
As the baby hears a word in her language
211
526000
3000
09:04
the auditory areas light up,
212
529000
2000
09:06
and then subsequently areas surrounding it
213
531000
2000
09:08
that we think are related to coherence,
214
533000
3000
09:11
getting the brain coordinated with its different areas,
215
536000
2000
09:13
and causality,
216
538000
2000
09:15
one brain area causing another to activate.
217
540000
3000
09:18
We are embarking
218
543000
2000
09:20
on a grand and golden age
219
545000
3000
09:23
of knowledge about child's brain development.
220
548000
3000
09:26
We're going to be able to see a child's brain
221
551000
2000
09:28
as they experience an emotion,
222
553000
2000
09:30
as they learn to speak and read,
223
555000
2000
09:32
as they solve a math problem,
224
557000
2000
09:34
as they have an idea.
225
559000
2000
09:36
And we're going to be able to invent brain-based interventions
226
561000
3000
09:39
for children who have difficulty learning.
227
564000
3000
09:42
Just as the poets and writers described,
228
567000
3000
09:45
we're going to be able to see, I think,
229
570000
2000
09:47
that wondrous openness,
230
572000
2000
09:49
utter and complete openness,
231
574000
2000
09:51
of the mind of a child.
232
576000
3000
09:54
In investigating the child's brain,
233
579000
2000
09:56
we're going to uncover deep truths
234
581000
2000
09:58
about what it means to be human,
235
583000
2000
10:00
and in the process,
236
585000
2000
10:02
we may be able to help keep our own minds open to learning
237
587000
2000
10:04
for our entire lives.
238
589000
2000
10:06
Thank you.
239
591000
2000
10:08
(Applause)
240
593000
3000
ABOUT THE SPEAKER
Patricia Kuhl - LanguagePatricia Kuhl studies how we learn language as babies, looking at the ways our brains form around language acquisition.
Why you should listen
Patricia Kuhl is co-director of the Institute for Brain and Learning Sciences at the University of Washington. She's internationally recognized for her research on early language and brain development, and studies that show how young children learn. Kuhl’s work has played a major role in demonstrating how early exposure to language alters the brain. It has implications for critical periods in development, for bilingual education and reading readiness, for developmental disabilities involving language, and for research on computer understanding of speech.
Patricia Kuhl | Speaker | TED.com