WEBVTT

00:00:00.040 --> 00:00:02.460
The following content is
provided under a Creative

00:00:02.460 --> 00:00:03.870
Commons license.

00:00:03.870 --> 00:00:06.910
Your support will help MIT
OpenCourseWare continue to

00:00:06.910 --> 00:00:10.560
offer high quality educational
resources for free.

00:00:10.560 --> 00:00:13.460
To make a donation or view
additional materials from

00:00:13.460 --> 00:00:19.290
hundreds of MIT courses, visit
MIT OpenCourseWare at

00:00:19.290 --> 00:00:20.540
ocw.mit.edu.

00:00:27.572 --> 00:00:29.470
PROFESSOR: OK, good afternoon.

00:00:29.470 --> 00:00:33.030
So in the last couple of weeks,
we've gone through some

00:00:33.030 --> 00:00:37.060
of the major human capacities
that you have and I have and

00:00:37.060 --> 00:00:38.840
most people have most
of them, right?

00:00:38.840 --> 00:00:43.530
Which are the ability to see the
world around us, to attend

00:00:43.530 --> 00:00:49.240
to what's important, to learn
lessons from life, to remember

00:00:49.240 --> 00:00:51.580
things, why we sometimes
forget things.

00:00:51.580 --> 00:00:53.880
And today, we're going to talk
about another huge dimension

00:00:53.880 --> 00:00:57.200
of human experience, of the
human mind, of the human

00:00:57.200 --> 00:00:58.940
brain, language.

00:00:58.940 --> 00:01:00.900
The ability to talk
to one another.

00:01:00.900 --> 00:01:02.900
I can talk, you can listen,
and vice versa.

00:01:02.900 --> 00:01:06.060
To read, to communicate,
in the ways that are so

00:01:06.060 --> 00:01:07.580
remarkable.

00:01:07.580 --> 00:01:10.270
Our life is flooded with
language, right?

00:01:10.270 --> 00:01:13.910
That we hear, that we
read, and so on.

00:01:13.910 --> 00:01:18.040
So when psychologists try to
divide up language and the

00:01:18.040 --> 00:01:20.610
study of both components, they
talk about comprehension,

00:01:20.610 --> 00:01:22.800
understanding what somebody
says to you when you're an

00:01:22.800 --> 00:01:27.900
infant, or now, as an adult, or
reading text, or producing

00:01:27.900 --> 00:01:30.720
speech, which you can do with
your mouth by speaking, or you

00:01:30.720 --> 00:01:31.440
learn to write.

00:01:31.440 --> 00:01:34.230
So you can both comprehend and
produce language, both

00:01:34.230 --> 00:01:36.340
visually and orally.

00:01:36.340 --> 00:01:41.720
And the thing that has greatly
enamored linguists to figure

00:01:41.720 --> 00:01:45.400
out is how language that we
speak is endlessly generative.

00:01:45.400 --> 00:01:47.980
I'm not sure you can actually
calculate out how many

00:01:47.980 --> 00:01:50.600
sentences we could possibly
produce, but just for fun,

00:01:50.600 --> 00:01:51.590
people have done stuff.

00:01:51.590 --> 00:01:53.870
And they said it would take 100
trillion years to memorize

00:01:53.870 --> 00:01:55.460
all the sentences
we can produce.

00:01:55.460 --> 00:01:56.890
I'm not sure that's
exactly the right

00:01:56.890 --> 00:01:58.540
number, but it's a lot.

00:01:58.540 --> 00:02:00.670
Because we can say a lot of
different sentences and

00:02:00.670 --> 00:02:02.030
understand what a person
is saying.

00:02:02.030 --> 00:02:05.970
It's an incredibly generative
system for communication, one

00:02:05.970 --> 00:02:08.090
person to another.

00:02:08.090 --> 00:02:12.120
So when people in psychology
think about different aspects

00:02:12.120 --> 00:02:14.590
of language, levels of language,
they often divide it

00:02:14.590 --> 00:02:16.190
into something like this.

00:02:16.190 --> 00:02:19.140
Phonology is the sounds of
language that we hear, that we

00:02:19.140 --> 00:02:22.700
acquire as infants, that we use
to speak to one another.

00:02:22.700 --> 00:02:24.490
Syntax, grammar.

00:02:24.490 --> 00:02:27.240
And by this, this isn't school
grammar, not get

00:02:27.240 --> 00:02:27.910
the sentence right.

00:02:27.910 --> 00:02:31.930
It's just the understanding of
the organization of words, and

00:02:31.930 --> 00:02:33.610
how that changes the meaning
of the sentence.

00:02:33.610 --> 00:02:34.930
Syntax.

00:02:34.930 --> 00:02:36.730
Semantics, what do
the words mean?

00:02:36.730 --> 00:02:38.060
What do the groups
of words mean?

00:02:38.060 --> 00:02:39.830
Pragmatics.

00:02:39.830 --> 00:02:42.150
And we'll talk about each
of these a little bit.

00:02:42.150 --> 00:02:46.420
Sometimes a person
is sarcastic.

00:02:46.420 --> 00:02:49.730
So they go, that was an awesome
lecture yesterday in

00:02:49.730 --> 00:02:50.240
psychology.

00:02:50.240 --> 00:02:53.190
Somebody could say that, in
a hundred trillion years.

00:02:53.190 --> 00:02:54.540
And you go, yeah, yeah.

00:02:54.540 --> 00:02:55.550
Right?

00:02:55.550 --> 00:02:58.490
So that's the pragmatics
versus semantics.

00:02:58.490 --> 00:03:00.700
Then, of course, it's
not just sentences.

00:03:00.700 --> 00:03:02.240
We have whole discussions.

00:03:02.240 --> 00:03:04.450
And another aspect of language
we'll just touch on a little

00:03:04.450 --> 00:03:07.760
bit today is emotional
comprehension and production.

00:03:07.760 --> 00:03:10.980
The tone somebody speaks in, the
tone you produce, are also

00:03:10.980 --> 00:03:11.270
informative.

00:03:11.270 --> 00:03:13.740
So there's many aspects
of language.

00:03:13.740 --> 00:03:15.890
We'll start with the most
basic, or phonology.

00:03:15.890 --> 00:03:18.320
So phonemes are the words that
people use to describe the

00:03:18.320 --> 00:03:20.350
building blocks of
speech sounds.

00:03:20.350 --> 00:03:23.010
So boy versus toy are different
by one phoneme.

00:03:23.010 --> 00:03:25.340
Phonemes are the building
blocks of

00:03:25.340 --> 00:03:27.030
the sounds of language.

00:03:27.030 --> 00:03:31.580
Humans use about 100 phonemes
across languages, but any

00:03:31.580 --> 00:03:34.540
given language only
uses a subset.

00:03:34.540 --> 00:03:36.530
And this has really interesting
consequences.

00:03:36.530 --> 00:03:40.020
Of course, a language could only
exist in the world if you

00:03:40.020 --> 00:03:41.970
could understand
those phonemes.

00:03:41.970 --> 00:03:43.435
Phonemes that are
ununderstandable aren't going

00:03:43.435 --> 00:03:45.030
to be a part of anybody's
language.

00:03:45.030 --> 00:03:47.820
But different languages, for
historical reasons, have ended

00:03:47.820 --> 00:03:50.520
up using different subsets
of phonemes.

00:03:50.520 --> 00:03:51.620
And we'll come back to that.

00:03:51.620 --> 00:03:54.480
When we're babies, for about
the first six months of our

00:03:54.480 --> 00:03:55.830
lives, we're universal
learners.

00:03:55.830 --> 00:03:58.440
We know all the phonemes and
respond to all of them.

00:03:58.440 --> 00:04:01.970
As we become native language
speakers and listeners, we

00:04:01.970 --> 00:04:05.120
become only good at the ones
that are used in our language,

00:04:05.120 --> 00:04:07.530
and our ability to even
recognize the ones that are

00:04:07.530 --> 00:04:09.920
not in our language
disappears.

00:04:09.920 --> 00:04:12.370
You can get it back some, but
the older that you are, you

00:04:12.370 --> 00:04:13.500
never get it back fully.

00:04:13.500 --> 00:04:17.060
So we're universal learners at
birth, but we become enslaved

00:04:17.060 --> 00:04:20.019
to the language that we are
native speakers of and listen

00:04:20.019 --> 00:04:21.720
to within months of birth.

00:04:25.430 --> 00:04:27.430
English has about 45.

00:04:27.430 --> 00:04:29.430
So it's more than the number
of letters we have.

00:04:29.430 --> 00:04:30.760
Some people get confused
a little bit

00:04:30.760 --> 00:04:31.920
that letters are phonemes.

00:04:31.920 --> 00:04:32.830
They're not.

00:04:32.830 --> 00:04:39.630
So hot and cold, the oh sound
in hot and the oh sound in

00:04:39.630 --> 00:04:41.540
cold are different sounds,
same letters.

00:04:41.540 --> 00:04:43.390
So there's many more phonemes.

00:04:43.390 --> 00:04:45.700
Even in English, when they say
there's about 45, depending on

00:04:45.700 --> 00:04:49.520
dialects, you could have a
couple more, a couple less.

00:04:49.520 --> 00:04:51.010
So again, we're born
to use all of them.

00:04:51.010 --> 00:04:52.290
Then we use them or lose them.

00:04:52.290 --> 00:04:54.430
It's something, we don't
entirely lose it, but we're

00:04:54.430 --> 00:04:58.150
never as good if we try to learn
a distinction once we

00:04:58.150 --> 00:04:59.980
become older than that.

00:04:59.980 --> 00:05:02.660
Humans understand speech
at an incredible rate.

00:05:02.660 --> 00:05:07.470
The typical fast talking is
about 180 words a minute, 14

00:05:07.470 --> 00:05:09.580
phonemes a second in a
continuous speech stream.

00:05:09.580 --> 00:05:11.560
But people can go
up to 250 words.

00:05:11.560 --> 00:05:16.530
We're incredibly fast language
processors, incredibly fast

00:05:16.530 --> 00:05:19.100
language processors.

00:05:19.100 --> 00:05:22.150
So how do we roughly
conceptualize, in the most

00:05:22.150 --> 00:05:25.780
simple way, what we have to do
to get to a word that we hear,

00:05:25.780 --> 00:05:26.950
to get to its meaning.

00:05:26.950 --> 00:05:28.910
Well, acoustic information
comes through the ear.

00:05:28.910 --> 00:05:33.850
Those are wave forms, physical
things, arriving at our ears.

00:05:33.850 --> 00:05:37.650
We in some way figure out how
those correspond to phonemes,

00:05:37.650 --> 00:05:40.260
and then map that in some
way to words that have

00:05:40.260 --> 00:05:41.510
meanings in the world.

00:05:44.450 --> 00:05:49.300
And the reason why machines
understanding what you say is

00:05:49.300 --> 00:05:52.200
shockingly hard to do
a good job on is for

00:05:52.200 --> 00:05:54.390
the following reason.

00:05:54.390 --> 00:05:57.400
There's a huge difference
between a word by the time

00:05:57.400 --> 00:06:00.360
your mind understands what it
is and the nature of the

00:06:00.360 --> 00:06:02.210
signal that arrived
at your year.

00:06:02.210 --> 00:06:03.770
So we talked a lot about
that in terms of

00:06:03.770 --> 00:06:05.440
the retina in vision.

00:06:05.440 --> 00:06:05.940
Right?

00:06:05.940 --> 00:06:08.170
You have this very impoverished
stuff that

00:06:08.170 --> 00:06:10.670
arrives in your retina, and then
the whole rest of your

00:06:10.670 --> 00:06:13.990
brain, millions and millions
of neurons and synapses,

00:06:13.990 --> 00:06:17.780
reconstruct what's out there
from a very impoverished

00:06:17.780 --> 00:06:19.110
retinal input.

00:06:19.110 --> 00:06:20.860
And the same thing happens
in audition.

00:06:20.860 --> 00:06:24.700
What gets to your ear is
incredibly impoverished, and

00:06:24.700 --> 00:06:27.790
the rest of your brain has to
brilliantly reconstruct what

00:06:27.790 --> 00:06:30.370
it heard and assign
its meaning.

00:06:30.370 --> 00:06:33.430
And so part of the challenge of
your mind and brain to do

00:06:33.430 --> 00:06:34.430
that is this.

00:06:34.430 --> 00:06:36.660
This is a representation of
something like the physical

00:06:36.660 --> 00:06:39.190
signal that arrives
at your ear.

00:06:39.190 --> 00:06:41.190
And you could say, well,
here's one word.

00:06:41.190 --> 00:06:41.850
Here's another word.

00:06:41.850 --> 00:06:43.260
What's the problem?

00:06:43.260 --> 00:06:46.410
But this is all the signal that
arrives at your ear for

00:06:46.410 --> 00:06:48.840
the word captain.

00:06:48.840 --> 00:06:54.450
So here's the C, all this stuff
for the A, a period of

00:06:54.450 --> 00:06:58.370
long silence that sets
up the P. Turns out--

00:06:58.370 --> 00:07:00.220
I'll convince you of
this in a moment--

00:07:00.220 --> 00:07:03.970
that our speech is full within
words of silent breaks that

00:07:03.970 --> 00:07:07.150
are incredibly informative and
alter our perception of what's

00:07:07.150 --> 00:07:09.630
physically going on.

00:07:09.630 --> 00:07:12.630
Then the little T. This
long stuff for the AI.

00:07:12.630 --> 00:07:14.200
This long stuff for the N.

00:07:14.200 --> 00:07:18.720
So the physical signal to your
ear barely resembles what

00:07:18.720 --> 00:07:21.190
seems like a single
unit, captain.

00:07:21.190 --> 00:07:22.740
And where's that pause?

00:07:22.740 --> 00:07:26.450
A person doesn't say, "cap," or
doesn't say "ca," one, two,

00:07:26.450 --> 00:07:29.380
three, "ptain." Right?

00:07:29.380 --> 00:07:31.630
I mean, but the pause is
happening, and it's incredibly

00:07:31.630 --> 00:07:32.690
informative, as I'll show you.

00:07:32.690 --> 00:07:34.890
So that's the incredible
thing.

00:07:34.890 --> 00:07:37.220
Your auditory system, your
language system, has to

00:07:37.220 --> 00:07:41.030
reconstruct sound into
phonemes, into words.

00:07:41.030 --> 00:07:45.760
It's not obvious in the physical
signal of the ear.

00:07:45.760 --> 00:07:48.340
It's not even that the physical
signal at the ear

00:07:48.340 --> 00:07:52.090
gives you much information at
the outset about whether

00:07:52.090 --> 00:07:54.680
you're hearing one word or
two words or three words.

00:07:54.680 --> 00:07:58.520
So here's "what," and "do you
mean" as one big blend.

00:07:58.520 --> 00:07:59.150
Where is "do"?

00:07:59.150 --> 00:07:59.850
Where is "you"?

00:07:59.850 --> 00:08:00.630
Where is "you mean"?

00:08:00.630 --> 00:08:02.930
And why does "what" look so
separate, and these three

00:08:02.930 --> 00:08:08.040
words, "what do you mean,"
get all garbled together.

00:08:08.040 --> 00:08:11.670
And you have to unpack them
by your brain and mind.

00:08:14.230 --> 00:08:15.870
Here's another example.

00:08:15.870 --> 00:08:18.060
John said that the dog
snapped at them.

00:08:18.060 --> 00:08:19.370
The dog snapped.

00:08:19.370 --> 00:08:22.000
Here's what's coming, roughly,
to your ear,

00:08:22.000 --> 00:08:23.310
taken from spoken speech.

00:08:23.310 --> 00:08:26.610
And here, again, notice that
the boundaries that you can

00:08:26.610 --> 00:08:31.380
see here, the gaps of silence,
occur not very related to the

00:08:31.380 --> 00:08:33.690
boundaries between the words
that your mind ultimately

00:08:33.690 --> 00:08:34.799
understands.

00:08:34.799 --> 00:08:37.740
So here's the word
"the," "dog".

00:08:37.740 --> 00:08:39.750
Here's S for "snapped".

00:08:39.750 --> 00:08:40.940
Here's a good pause.

00:08:40.940 --> 00:08:43.809
But that's a pause between the
P sound and the T sound.

00:08:43.809 --> 00:08:45.650
That looks like a good
word boundary.

00:08:45.650 --> 00:08:48.750
To a machine, that's a very
good word boundary.

00:08:48.750 --> 00:08:50.440
It's not a word boundary
for you.

00:08:50.440 --> 00:08:52.290
You know snapped is one word.

00:08:52.290 --> 00:08:56.090
And here comes "at." "At" is
kind of blended into the T. So

00:08:56.090 --> 00:08:59.090
there's this physical signal
doesn't even give you breaks

00:08:59.090 --> 00:09:01.800
where words separate
from one another.

00:09:01.800 --> 00:09:07.230
It's kind of a continuous mix
of breaks and silences.

00:09:07.230 --> 00:09:12.090
So here's to convince you that
a silence carries incredible

00:09:12.090 --> 00:09:14.560
information in it.

00:09:14.560 --> 00:09:16.050
Let's see, make sure
I have the volume.

00:09:16.050 --> 00:09:18.570
OK, so I'm going to
play you this.

00:09:18.570 --> 00:09:23.850
This is a physical signal to
your ear of the word "say."

00:09:23.850 --> 00:09:24.970
RECORDING: Say.

00:09:24.970 --> 00:09:27.330
PROFESSOR: OK.

00:09:27.330 --> 00:09:29.670
Yeah?

00:09:29.670 --> 00:09:31.460
I always worry because Tyler's
an expert in this, and I'm

00:09:31.460 --> 00:09:33.250
always like, please,
Tyler, don't

00:09:33.250 --> 00:09:35.310
embarrass me when I'm wrong.

00:09:35.310 --> 00:09:37.520
You don't have this pressure
out there, but the teaching

00:09:37.520 --> 00:09:39.270
assistants are all experts
in something or another.

00:09:39.270 --> 00:09:41.690
And then they know more often
than I do on that topic, and

00:09:41.690 --> 00:09:42.860
we hit their topic
in the course.

00:09:42.860 --> 00:09:45.910
I'm looking over there
nervously.

00:09:45.910 --> 00:09:49.150
So here's "say." Now, they're
going to take this thing and

00:09:49.150 --> 00:09:51.980
just separate this, just in
the way you do a physical

00:09:51.980 --> 00:09:53.450
signal on a computer, just
separate in time.

00:09:53.450 --> 00:09:55.290
They're just going to add
a little bit of time.

00:09:55.290 --> 00:09:58.280
And what do you hear?

00:09:58.280 --> 00:09:59.880
RECORDING: Stay.

00:09:59.880 --> 00:10:04.390
PROFESSOR: The T sound that
makes is simply adding silence

00:10:04.390 --> 00:10:06.010
here to produce that.

00:10:06.010 --> 00:10:07.140
Now, that's what you
hear all the time.

00:10:07.140 --> 00:10:08.380
It's not a random choice.

00:10:08.380 --> 00:10:09.470
But isn't that amazing?

00:10:09.470 --> 00:10:13.240
That silence is as ultimately
informative for your mind as

00:10:13.240 --> 00:10:16.060
actual language sounds
themselves.

00:10:16.060 --> 00:10:18.200
All this is happening on the
order of thousandths of a

00:10:18.200 --> 00:10:19.595
second, in bits of
information.

00:10:22.130 --> 00:10:22.870
That's bad enough.

00:10:22.870 --> 00:10:24.670
Here's another amazingly
tough thing.

00:10:24.670 --> 00:10:28.440
So people like to compare
consonants that have a lot of

00:10:28.440 --> 00:10:31.060
similarities with some
interesting differences.

00:10:31.060 --> 00:10:31.640
Ba and da.

00:10:31.640 --> 00:10:33.940
This is really comparing B
versus D. It's throwing on the

00:10:33.940 --> 00:10:35.650
A just to have a vowel sound.

00:10:35.650 --> 00:10:38.260
So here, again, is a
representation of what's

00:10:38.260 --> 00:10:39.820
coming to your ear.

00:10:39.820 --> 00:10:42.580
Notice that all the difference
between the B and the D, these

00:10:42.580 --> 00:10:45.440
two consonants, all the physical
differences occurring

00:10:45.440 --> 00:10:48.070
in 40,000ths of a second.

00:10:48.070 --> 00:10:51.210
40,000ths of a second is all the
physical difference that

00:10:51.210 --> 00:10:52.990
you get as you hear
these two things.

00:10:56.720 --> 00:10:59.660
RECORDING: Ba.

00:10:59.660 --> 00:11:00.440
Da.

00:11:00.440 --> 00:11:03.530
PROFESSOR: 40,000ths of a
second, and yet you could hear

00:11:03.530 --> 00:11:06.980
the word bad, and you're
not going bad, dab.

00:11:06.980 --> 00:11:08.790
Come on, that was thousandths
of a second, right?

00:11:08.790 --> 00:11:11.220
And not only that, there'd be
another word and another word

00:11:11.220 --> 00:11:11.860
in a sentence.

00:11:11.860 --> 00:11:14.490
That's why people use the word
speech stream, and we like to

00:11:14.490 --> 00:11:16.970
think of it as a
speech torrent.

00:11:16.970 --> 00:11:20.610
Phonemes, phonemes, phonemes,
breaks, all garbled together

00:11:20.610 --> 00:11:23.480
in your mind as you back and
constantly, instantly, and

00:11:23.480 --> 00:11:27.050
brilliantly figure out
what you've heard.

00:11:27.050 --> 00:11:29.910
So let me give you another
example because part of the

00:11:29.910 --> 00:11:34.630
challenge of this are phonemes
occur in the context of words

00:11:34.630 --> 00:11:35.370
and sentences.

00:11:35.370 --> 00:11:37.800
So make sure I have good
volume on this.

00:11:37.800 --> 00:11:40.080
OK, so let's do this
first one.

00:11:40.080 --> 00:11:41.130
RECORDING: [INAUDIBLE].

00:11:41.130 --> 00:11:42.334
PROFESSOR: What was that?

00:11:42.334 --> 00:11:42.806
AUDIENCE: [INAUDIBLE].

00:11:42.806 --> 00:11:43.278
PROFESSOR: What?

00:11:43.278 --> 00:11:44.700
AUDIENCE: [INAUDIBLE].

00:11:44.700 --> 00:11:47.510
PROFESSOR: OK.

00:11:47.510 --> 00:11:50.700
Here's the same thing, and a
person is going to say this

00:11:50.700 --> 00:11:52.790
now all by itself.

00:11:52.790 --> 00:11:53.700
RECORDING: Little.

00:11:53.700 --> 00:11:54.680
PROFESSOR: You got
that, right?

00:11:54.680 --> 00:11:56.526
OK.

00:11:56.526 --> 00:11:59.040
RECORDING: Mary saw the three
little pigs in the corridor.

00:11:59.040 --> 00:12:01.460
PROFESSOR: OK, but here's
the amazing thing.

00:12:01.460 --> 00:12:04.990
The first sound you heard, and
I'll replay it, they took the

00:12:04.990 --> 00:12:08.170
word little, and they cut
it out from the last

00:12:08.170 --> 00:12:10.020
sentence you heard.

00:12:10.020 --> 00:12:13.490
When little is said all by
itself, what comes to your ear

00:12:13.490 --> 00:12:15.730
is literally the first
thing you heard.

00:12:15.730 --> 00:12:20.120
Not this, because when you say
a word in isolation, little,

00:12:20.120 --> 00:12:21.490
people say it differently.

00:12:21.490 --> 00:12:24.080
But we almost never
talk like that.

00:12:24.080 --> 00:12:26.110
We almost always have
a sentence.

00:12:26.110 --> 00:12:29.200
And it turns out the way we say
things, and therefore, the

00:12:29.200 --> 00:12:33.130
way you hear things, is hugely
influenced by what comes

00:12:33.130 --> 00:12:35.910
before it and what
comes after it.

00:12:35.910 --> 00:12:37.810
Never mind things like the
speed at which different

00:12:37.810 --> 00:12:40.920
people talk, the accent they
talk, or the loudness of the

00:12:40.920 --> 00:12:42.810
background conversation.

00:12:42.810 --> 00:12:45.286
But just simply the fact
that you're going to

00:12:45.286 --> 00:12:46.700
say little by itself.

00:12:46.700 --> 00:12:47.500
How does that work?

00:12:47.500 --> 00:12:48.620
It's amazing, right?

00:12:48.620 --> 00:12:49.340
Listen again.

00:12:49.340 --> 00:12:52.050
If you heard little all by
itself, and it's exactly the

00:12:52.050 --> 00:12:53.300
same physical signal.

00:12:56.834 --> 00:12:57.302
Maybe it's not so--

00:12:57.302 --> 00:12:58.710
RECORDING: [INAUDIBLE].

00:12:58.710 --> 00:12:59.720
PROFESSOR: OK?

00:12:59.720 --> 00:13:01.132
Let me try that again.

00:13:01.132 --> 00:13:03.180
RECORDING: [INAUDIBLE].

00:13:03.180 --> 00:13:06.350
PROFESSOR: That's exactly the
same physical signal, exactly

00:13:06.350 --> 00:13:07.330
the same physical signal, as--

00:13:07.330 --> 00:13:10.180
RECORDING: Mary saw the three
little pigs in the corridor.

00:13:10.180 --> 00:13:11.650
PROFESSOR: What you
hear easily as

00:13:11.650 --> 00:13:12.850
little is exactly that.

00:13:12.850 --> 00:13:13.910
Words in isolation--

00:13:13.910 --> 00:13:15.250
and there's a great
demonstration.

00:13:15.250 --> 00:13:18.900
If you hear a word cut out, the
reason why it sounds like

00:13:18.900 --> 00:13:22.710
nothing is your mind is already
preparing how to

00:13:22.710 --> 00:13:25.790
interpret that, given what you
heard just before it, and also

00:13:25.790 --> 00:13:28.080
a little bit what you
hear just after it.

00:13:28.080 --> 00:13:31.370
Your mind is constantly fixing
up things because it knows

00:13:31.370 --> 00:13:33.240
what's going on, what's going
into it, what's coming out.

00:13:33.240 --> 00:13:36.240
It's not because it just simply
hears the word itself.

00:13:36.240 --> 00:13:38.480
If we lived in a world where
people just say one word every

00:13:38.480 --> 00:13:41.270
half an hour, we
could do this.

00:13:41.270 --> 00:13:43.320
In a world where we're talking
to each other in sentences all

00:13:43.320 --> 00:13:45.190
the time, we have to do this.

00:13:47.870 --> 00:13:51.030
Here's another example of our
brilliant auditory system for

00:13:51.030 --> 00:13:52.280
hearing these kinds of things.

00:13:52.280 --> 00:13:54.620
So again, these are
representations of how this

00:13:54.620 --> 00:13:57.040
physical signal arrives
at your ear.

00:13:57.040 --> 00:14:01.050
Here's the word bet, bee,
boat, and bird.

00:14:01.050 --> 00:14:03.390
They all start with
the same B sound.

00:14:03.390 --> 00:14:04.350
That's easy, right?

00:14:04.350 --> 00:14:06.020
But look what came
to your ear.

00:14:06.020 --> 00:14:09.350
Look at this signal, this
signal, this signal.

00:14:09.350 --> 00:14:09.950
Look at this one.

00:14:09.950 --> 00:14:11.950
How different does this look?

00:14:11.950 --> 00:14:15.900
It's different because what
follows bird influences how

00:14:15.900 --> 00:14:19.250
this is said, but you can see
that you can translate both of

00:14:19.250 --> 00:14:21.440
these into the B sound
pretty easily.

00:14:21.440 --> 00:14:22.260
Now, look at these words.

00:14:22.260 --> 00:14:25.680
These all start with D. Look
at how bet and debt look

00:14:25.680 --> 00:14:26.540
almost identical.

00:14:26.540 --> 00:14:29.000
We said it's only 40
milliseconds difference,

00:14:29.000 --> 00:14:31.050
40,000ths of a second difference
that physically

00:14:31.050 --> 00:14:32.080
moved through your ear.

00:14:32.080 --> 00:14:34.920
But you can tell the word
bet versus debt.

00:14:34.920 --> 00:14:37.030
Look at the word D, how
different it looks

00:14:37.030 --> 00:14:38.840
from word to word.

00:14:38.840 --> 00:14:40.720
So your mind is constantly
figuring

00:14:40.720 --> 00:14:42.120
out what you're hearing.

00:14:42.120 --> 00:14:45.370
The problem of invariance, and
it's constantly coming to you

00:14:45.370 --> 00:14:47.650
in incredibly variable ways.

00:14:51.170 --> 00:14:54.550
So one of the tools we have, and
an amazing human capacity,

00:14:54.550 --> 00:14:56.820
is what's called categorical
perception.

00:14:56.820 --> 00:14:59.780
So we have all this huge range
of physical sounds that can

00:14:59.780 --> 00:15:01.050
come to our ear.

00:15:01.050 --> 00:15:02.650
And kind of brilliantly,
and you don't

00:15:02.650 --> 00:15:03.480
even think about it--

00:15:03.480 --> 00:15:05.170
I mean, none of you go around
bragging about your

00:15:05.170 --> 00:15:06.900
categorical perception
abilities, right?

00:15:06.900 --> 00:15:09.070
Like when you go to spring
break, right?

00:15:09.070 --> 00:15:11.340
And maybe some of you will go
home, or you'll see friends,

00:15:11.340 --> 00:15:13.740
and you won't go, I have
categorical perception.

00:15:13.740 --> 00:15:16.600
I try to be modest about it.

00:15:16.600 --> 00:15:21.290
But it's an amazing human
talent, and essential because

00:15:21.290 --> 00:15:22.420
what it does is this.

00:15:22.420 --> 00:15:25.310
We hear this huge range of
sounds, depending on accents,

00:15:25.310 --> 00:15:27.040
what comes before,
what comes after.

00:15:27.040 --> 00:15:29.530
And we have to decide that's a
B. And what you're going to do

00:15:29.530 --> 00:15:32.270
is take this huge range of
things, and just go, boom.

00:15:32.270 --> 00:15:35.360
It's a B or a D, and that's
all there is to it.

00:15:35.360 --> 00:15:38.290
So here's the way they do it.

00:15:38.290 --> 00:15:39.890
So categorical perception.

00:15:39.890 --> 00:15:41.480
Sounds can vary continuously.

00:15:41.480 --> 00:15:43.460
We showed you some of
this continuity.

00:15:43.460 --> 00:15:45.060
In this experiment, on purpose,
they're going to vary

00:15:45.060 --> 00:15:48.420
sounds in a very continuous,
physical way.

00:15:48.420 --> 00:15:53.800
But in our heads, for ba and da
here, in our heads, all the

00:15:53.800 --> 00:15:56.290
stuff that's over here, even
though they're very different

00:15:56.290 --> 00:15:58.690
one from the other at a
physical, specific level,

00:15:58.690 --> 00:15:59.700
we'll say, ba.

00:15:59.700 --> 00:16:00.360
I'm done with it.

00:16:00.360 --> 00:16:03.250
We're going to put it in one
category for interpretation.

00:16:03.250 --> 00:16:05.750
And you're going to have a
pretty sharp boundary here.

00:16:05.750 --> 00:16:08.510
And then everything over here,
all these physical signals,

00:16:08.510 --> 00:16:09.780
you're going to call pa.

00:16:09.780 --> 00:16:12.020
And so you're not going to spend
a lot of time going,

00:16:12.020 --> 00:16:15.720
that was a one third ba,
2/3 pa, probability.

00:16:15.720 --> 00:16:16.690
You couldn't do that, right?

00:16:16.690 --> 00:16:17.690
Because you couldn't
get through a

00:16:17.690 --> 00:16:18.680
word, nevermind a sentence.

00:16:18.680 --> 00:16:21.760
You're going to instantly group
huge numbers of things

00:16:21.760 --> 00:16:24.520
that are different from one
another into two categories

00:16:24.520 --> 00:16:26.470
that need to be separated.

00:16:26.470 --> 00:16:28.360
So here's a kind of experiment
they do.

00:16:28.360 --> 00:16:29.670
I'll just give you the
feeling of it.

00:16:29.670 --> 00:16:32.070
Here's what they're varying,
constant units.

00:16:32.070 --> 00:16:35.350
Constant units, every stimulus
here is different.

00:16:35.350 --> 00:16:38.335
But every time you hear ba, and
they're constantly going

00:16:38.335 --> 00:16:40.900
to vary something by its
physical property, you always

00:16:40.900 --> 00:16:45.910
call it a B. And then something
happens, in about 20

00:16:45.910 --> 00:16:49.360
milliseconds of information, and
boom, you always call it a

00:16:49.360 --> 00:16:53.070
P. So change, change, change,
change, change, in the

00:16:53.070 --> 00:16:56.340
physical signal, you call
it one category.

00:16:56.340 --> 00:16:58.150
A little bit more
change, boom.

00:16:58.150 --> 00:16:59.540
It's the other category.

00:16:59.540 --> 00:17:01.940
And you keep changing, but for
you, it's always a perfectly

00:17:01.940 --> 00:17:06.079
good P, even though every one
of those are different.

00:17:06.079 --> 00:17:10.310
So it's an amazing fundamental
capacity to take a huge, and,

00:17:10.310 --> 00:17:13.260
for all practical purposes, an
infinite variety of inputs and

00:17:13.260 --> 00:17:17.520
say it's a B or a P if that's
the world I'm working in.

00:17:17.520 --> 00:17:19.190
Syntax.

00:17:19.190 --> 00:17:21.619
That's another way in which we
understand things, the way

00:17:21.619 --> 00:17:23.010
that words are organized
and the order.

00:17:23.010 --> 00:17:25.839
So we could say, "The model
embraced the designer and the

00:17:25.839 --> 00:17:27.569
photographer." Imagine somebody
saying to this.

00:17:27.569 --> 00:17:30.240
But then they complete the
sentence, "The model embraced

00:17:30.240 --> 00:17:33.070
the designer and the
photographer laughed."

00:17:33.070 --> 00:17:35.890
So people who study syntax
really are interested in what

00:17:35.890 --> 00:17:38.180
happens in your mind as you
hear, "The model embraced the

00:17:38.180 --> 00:17:40.600
designer and the photographer,"
and some big

00:17:40.600 --> 00:17:42.420
group hug is going on, right?

00:17:42.420 --> 00:17:45.130
And then all of a sudden, you
get this word laughed, and you

00:17:45.130 --> 00:17:47.210
have to rearrange the elements
in your mind.

00:17:47.210 --> 00:17:49.020
That's what syntax is letting
you do, to say, oh, the

00:17:49.020 --> 00:17:50.620
meaning is really
quite different.

00:17:50.620 --> 00:17:53.560
There's only one hug going on,
and one person laughing at

00:17:53.560 --> 00:17:54.210
them, right?

00:17:54.210 --> 00:17:58.090
But the meaning over time as
somebody talks to you set up,

00:17:58.090 --> 00:18:00.030
you're thinking this, boom.

00:18:00.030 --> 00:18:05.150
Because you understand syntax,
you understand what happens.

00:18:05.150 --> 00:18:07.870
Semantics are kind of more
straightforward than how we

00:18:07.870 --> 00:18:09.190
usually think about them.

00:18:09.190 --> 00:18:12.880
They're the meaning of words
or sentences, morphemes.

00:18:12.880 --> 00:18:14.640
And we can understand semantics
even when the

00:18:14.640 --> 00:18:16.850
sentence is nonsense. "Colorless
green ideas sleep

00:18:16.850 --> 00:18:19.330
furiously" is a famous
example.

00:18:19.330 --> 00:18:21.930
We can understand semantics
even in bad grammatical

00:18:21.930 --> 00:18:23.020
sentences, right?

00:18:23.020 --> 00:18:27.050
"Fastly eat dinner, ballgame
start soon." Your fifth grade

00:18:27.050 --> 00:18:31.190
teacher would be up in arms to
hear that, but you understand.

00:18:31.190 --> 00:18:31.660
Eat now.

00:18:31.660 --> 00:18:33.690
Go see the ball game.

00:18:33.690 --> 00:18:36.390
So we get through a lot of
stuff to get semantics.

00:18:36.390 --> 00:18:37.460
And we can see interactions.

00:18:37.460 --> 00:18:40.820
I stole this example from Ted
Gibson because it took me 20

00:18:40.820 --> 00:18:43.170
minutes to figure out what
the sentence means.

00:18:43.170 --> 00:18:44.420
You're probably faster
than I am.

00:18:48.791 --> 00:18:50.600
Are you OK?

00:18:50.600 --> 00:18:52.060
I'm like the old man.

00:18:52.060 --> 00:18:53.820
Where's the missing words?

00:18:53.820 --> 00:18:56.730
The old man the boats.

00:18:56.730 --> 00:18:59.270
So this is how you have to
parse up the sentence.

00:18:59.270 --> 00:19:00.450
It's not always obvious.

00:19:00.450 --> 00:19:02.240
People who study language love
these things, where people are

00:19:02.240 --> 00:19:03.580
all confused.

00:19:03.580 --> 00:19:05.320
Most of our lives, we don't go
through that much confusion.

00:19:05.320 --> 00:19:07.470
But there's some fun ones,
sort of like the visual

00:19:07.470 --> 00:19:10.340
illusions, to point out how
meaning and syntax can

00:19:10.340 --> 00:19:11.590
interact as you figure
out things.

00:19:11.590 --> 00:19:12.920
Here's another example.

00:19:12.920 --> 00:19:15.330
Jay Leno talked about sex
with Lindsey Lohan.

00:19:15.330 --> 00:19:17.350
OK, now wait a minute.

00:19:17.350 --> 00:19:21.845
Did I miss that excellent
episode of Jay Leno, or was

00:19:21.845 --> 00:19:23.730
she on the show, and talking
about-- she was saying I'd

00:19:23.730 --> 00:19:25.480
rather not talk about my
recent conviction.

00:19:25.480 --> 00:19:29.170
I'd rather talk about something
less controversial.

00:19:29.170 --> 00:19:30.720
So semantic.

00:19:30.720 --> 00:19:31.900
All these things are constantly

00:19:31.900 --> 00:19:32.810
interacting in your mind.

00:19:32.810 --> 00:19:34.740
It's kind of amazing when you
think about that, right?

00:19:34.740 --> 00:19:36.090
We said phonology's amazing.

00:19:36.090 --> 00:19:37.910
Semantics, syntax, boom.

00:19:37.910 --> 00:19:39.050
They're all happening in you.

00:19:39.050 --> 00:19:40.220
And none of you are
sitting there,

00:19:40.220 --> 00:19:42.830
going, just do phonology.

00:19:42.830 --> 00:19:44.570
Please don't ask me
to do semantics.

00:19:44.570 --> 00:19:45.910
We just had midterms, right?

00:19:45.910 --> 00:19:48.570
Brilliant.

00:19:48.570 --> 00:19:50.780
So one of the ways that people
try to convince themselves

00:19:50.780 --> 00:19:52.530
that these things are separable
and gives a little

00:19:52.530 --> 00:19:54.990
bit more insight is Evoked
Response Potentials.

00:19:54.990 --> 00:19:57.560
Those are these things that you
can measure on infants or

00:19:57.560 --> 00:20:00.910
adults that measure electrical
activity in milliseconds and

00:20:00.910 --> 00:20:04.610
the speed of language
processing, surface electrodes

00:20:04.610 --> 00:20:10.260
that give you an index of a few
hundred thousand neurons.

00:20:10.260 --> 00:20:13.280
And this is an example you saw
before, actually, earlier in

00:20:13.280 --> 00:20:15.280
the course, the so-called
N400.

00:20:15.280 --> 00:20:18.590
So N means whether the
wave is going up or

00:20:18.590 --> 00:20:20.900
down, negative or positive.

00:20:20.900 --> 00:20:24.200
400 means that it occurs about
400 milliseconds, about half a

00:20:24.200 --> 00:20:26.410
second, after the
relevant event.

00:20:26.410 --> 00:20:31.380
So here is this, let's see, it
was his first day at work.

00:20:31.380 --> 00:20:32.280
That's the baseline.

00:20:32.280 --> 00:20:34.550
He spread the warm
bread with socks.

00:20:34.550 --> 00:20:35.360
Oh my gosh, socks.

00:20:35.360 --> 00:20:36.160
Semantically bad.

00:20:36.160 --> 00:20:38.760
And here comes the N400.

00:20:38.760 --> 00:20:42.080
And the control condition is you
throw up this unexpected

00:20:42.080 --> 00:20:45.900
word, shoes, that's weird, but
not semantically, and you get

00:20:45.900 --> 00:20:46.890
quite a different response.

00:20:46.890 --> 00:20:50.520
So here's an N400 that seems
to simply be about semantic

00:20:50.520 --> 00:20:53.690
expectation or congruence.

00:20:53.690 --> 00:20:57.110
How about if you
mess up syntax?

00:20:57.110 --> 00:21:02.780
The spoiled child throw
the toys on the box.

00:21:02.780 --> 00:21:06.460
Then you get a response at
about 600 milliseconds to

00:21:06.460 --> 00:21:08.270
syntactic incongruity.

00:21:08.270 --> 00:21:11.740
So people, this is just more
evidence that in the brain,

00:21:11.740 --> 00:21:14.720
meaning or semantics and syntax
or something about word

00:21:14.720 --> 00:21:20.470
order are processed somewhat
differently in time and place.

00:21:20.470 --> 00:21:21.600
Pragmatics is kind of fun.

00:21:21.600 --> 00:21:23.410
It's practical understanding.

00:21:23.410 --> 00:21:24.660
The old joke, do you know
what time it is?

00:21:24.660 --> 00:21:25.990
And the person says, yes.

00:21:25.990 --> 00:21:27.705
They're messing with your
sense of pragmatics.

00:21:30.380 --> 00:21:31.800
Two negatives make a positive.

00:21:31.800 --> 00:21:32.590
Saying, yeah, yeah.

00:21:32.590 --> 00:21:35.750
I've heard that one before,
earlier in the lecture.

00:21:35.750 --> 00:21:37.860
Now, here's something kind of
interesting, again, which is

00:21:37.860 --> 00:21:38.195
remarkable.

00:21:38.195 --> 00:21:40.920
And if you didn't do the
experiment, you couldn't know

00:21:40.920 --> 00:21:42.450
that our mind is like this.

00:21:42.450 --> 00:21:45.370
Because you wouldn't do things
like this because it sounds

00:21:45.370 --> 00:21:48.090
too ridiculous, but your mind
does it all the time.

00:21:48.090 --> 00:21:51.300
So we're going to think about
the word "bugs." And I have to

00:21:51.300 --> 00:21:53.570
set you up a little bit with
paradigms you've heard before,

00:21:53.570 --> 00:21:55.490
but just remind you
of these things.

00:21:55.490 --> 00:21:59.660
So one kind of task that
psychologists like to use to

00:21:59.660 --> 00:22:01.410
think about how people
process words is a

00:22:01.410 --> 00:22:03.850
lexical decision task.

00:22:03.850 --> 00:22:07.790
All you're deciding is whether
a word is a real word or a

00:22:07.790 --> 00:22:09.340
nonsense word.

00:22:09.340 --> 00:22:10.580
So doctor is real.

00:22:10.580 --> 00:22:11.530
Poctor is not.

00:22:11.530 --> 00:22:12.950
Spy is a real word.

00:22:12.950 --> 00:22:15.250
This spelling is not
a real word.

00:22:15.250 --> 00:22:17.700
So real word, not a real word.

00:22:17.700 --> 00:22:20.420
We also talked about one other
idea early in the course, that

00:22:20.420 --> 00:22:24.440
if, prior to processing a word
like doctor, you get an

00:22:24.440 --> 00:22:27.500
unrelated word or a related
word, you process doctor

00:22:27.500 --> 00:22:30.120
faster if you had a related
word before it.

00:22:30.120 --> 00:22:30.870
Semantic priming.

00:22:30.870 --> 00:22:33.260
The ideas about nurse
overlap a lot with

00:22:33.260 --> 00:22:34.840
the ideas about doctor.

00:22:34.840 --> 00:22:36.150
And seeing this word
before makes you

00:22:36.150 --> 00:22:37.810
process this word faster.

00:22:37.810 --> 00:22:39.490
So you need those two
things in mind.

00:22:39.490 --> 00:22:40.730
What was the person doing?

00:22:40.730 --> 00:22:42.110
Is it a real word or not?

00:22:42.110 --> 00:22:45.060
And you're going to mess with
what's in their mind before

00:22:45.060 --> 00:22:46.980
that, related or unrelated.

00:22:46.980 --> 00:22:49.280
And here's the experiment.

00:22:49.280 --> 00:22:51.750
So you would hear, over
headphones, something like

00:22:51.750 --> 00:22:52.730
this, as an example.

00:22:52.730 --> 00:22:55.760
"Rumor has it that for years
the government building had

00:22:55.760 --> 00:22:56.910
been plagued with
problems." Now,

00:22:56.910 --> 00:22:58.510
here's a critical sentence.

00:22:58.510 --> 00:23:01.560
"The man was not surprised when
he found several spiders,

00:23:01.560 --> 00:23:05.330
roaches, and other bugs in
the corner of his room."

00:23:05.330 --> 00:23:07.550
So there's two senses of the
word, at least two meanings of

00:23:07.550 --> 00:23:08.740
the word bug, right?

00:23:08.740 --> 00:23:12.870
The insect you don't want in
your soup, or the microphone

00:23:12.870 --> 00:23:14.490
that somebody slipped into
your room to listen to

00:23:14.490 --> 00:23:15.210
what's going on.

00:23:15.210 --> 00:23:18.350
Two different meanings
for the word bug.

00:23:18.350 --> 00:23:20.800
This is obviously about
the insect.

00:23:20.800 --> 00:23:23.210
So the person is hearing this
over headphones, and on a

00:23:23.210 --> 00:23:27.150
computer monitor in front of
them, after they hear the word

00:23:27.150 --> 00:23:31.600
bugs here, they see, for a
lexical decision, the word

00:23:31.600 --> 00:23:35.870
"ant," which is related to this
meaning, or "spy," which

00:23:35.870 --> 00:23:38.120
is related to the other meaning
that's not relevant

00:23:38.120 --> 00:23:41.110
for here, or "sew," which is
the control condition.

00:23:41.110 --> 00:23:42.990
We'll ignore that because it's
just a control condition.

00:23:42.990 --> 00:23:44.140
I'll ignore that for now.

00:23:44.140 --> 00:23:47.220
So you're making a lexical
decision now on the related

00:23:47.220 --> 00:23:49.680
meaning or the unrelated
meaning.

00:23:49.680 --> 00:23:51.760
And here's what they find.

00:23:51.760 --> 00:23:55.200
If it's the related meaning, and
you only have about a half

00:23:55.200 --> 00:23:57.960
second from when that came on
to when you had to do this

00:23:57.960 --> 00:23:59.770
task, so you saw the word
bug and here comes

00:23:59.770 --> 00:24:03.290
this, you're faster.

00:24:03.290 --> 00:24:07.100
But if you get the unrelated
meaning within a half a second

00:24:07.100 --> 00:24:11.240
of processing the word,
you're also faster.

00:24:11.240 --> 00:24:14.280
What this means is, your mind
has done exactly what you

00:24:14.280 --> 00:24:15.800
would think it wouldn't do.

00:24:15.800 --> 00:24:18.520
Your mind has looked up in
your brain, and in your

00:24:18.520 --> 00:24:22.720
knowledge of vocabulary, every
meaning of the word.

00:24:22.720 --> 00:24:25.310
Even though the sentence is
clearly setting you up for the

00:24:25.310 --> 00:24:30.010
word, the insect meaning, every
meaning of your word is

00:24:30.010 --> 00:24:31.440
turned on and available
to you.

00:24:31.440 --> 00:24:33.290
Your mind does the opposite
of what you might think.

00:24:33.290 --> 00:24:34.820
You might think, well, I know--

00:24:34.820 --> 00:24:36.920
let's look at the sentence.

00:24:36.920 --> 00:24:39.460
I know, once I get spiders,
roaches, and bugs.

00:24:39.460 --> 00:24:42.010
By here, you know you're in the
insect world, not in the

00:24:42.010 --> 00:24:43.950
world of spies, right?

00:24:43.950 --> 00:24:50.170
But your mind, even then, looks
up every possible word.

00:24:50.170 --> 00:24:53.840
Because it's cheaper for the
mind to look up every word

00:24:53.840 --> 00:24:57.360
than to start to decide what
the appropriate word is.

00:24:57.360 --> 00:25:01.780
And then, if you wait another
second and a half, you're only

00:25:01.780 --> 00:25:04.900
fast for the selected meaning.

00:25:04.900 --> 00:25:06.510
And now you're slowed
down for the one

00:25:06.510 --> 00:25:08.100
that's not context relevant.

00:25:08.100 --> 00:25:11.030
So your mind looks up everything
all the time.

00:25:11.030 --> 00:25:13.310
Every meaning of every
word you know.

00:25:13.310 --> 00:25:15.320
Every time that word pops up, it
seems like your mind looks

00:25:15.320 --> 00:25:18.650
up every meaning, and then the
rest of your mind selects out

00:25:18.650 --> 00:25:20.720
the one that's relevant
and shuts down the

00:25:20.720 --> 00:25:22.150
ones that are not.

00:25:22.150 --> 00:25:22.900
But you wouldn't think that.

00:25:22.900 --> 00:25:24.280
You would think, maybe I'll
just get the one I need.

00:25:24.280 --> 00:25:24.955
I'm not going to
get everything.

00:25:24.955 --> 00:25:27.110
It's like, I'm not going to put
all my clothes on, walk

00:25:27.110 --> 00:25:28.960
outside, and take off everything
that's not relevant

00:25:28.960 --> 00:25:30.620
for the weather, right?

00:25:30.620 --> 00:25:34.070
It seems like a really
inefficient way to get

00:25:34.070 --> 00:25:34.970
dressed, right?

00:25:34.970 --> 00:25:36.760
But that's not the
way your mind is.

00:25:36.760 --> 00:25:39.950
Your mind is cheaper and more
efficient to look up every

00:25:39.950 --> 00:25:42.410
meaning every time, and then
pick the right one.

00:25:42.410 --> 00:25:45.400
It's more efficient, and that's
how people do it.

00:25:45.400 --> 00:25:47.850
So people use this word
exhaustive lexical access.

00:25:47.850 --> 00:25:50.670
I look up every meaning of a
word, and then I'll pick out

00:25:50.670 --> 00:25:51.920
the correct one.

00:25:56.720 --> 00:26:00.920
Another interesting aspect that
could be brought into

00:26:00.920 --> 00:26:02.700
pragmatics, and we'll talk a
little bit more about this

00:26:02.700 --> 00:26:04.860
later in the course, but
emotional intonation.

00:26:04.860 --> 00:26:07.910
And it turns out where it's
almost all language processes

00:26:07.910 --> 00:26:09.620
are left dominance as
we talked about many

00:26:09.620 --> 00:26:11.140
times in the course.

00:26:11.140 --> 00:26:15.980
One stream of processing
language, both visual and

00:26:15.980 --> 00:26:18.570
auditory, is right hemisphere
dependent.

00:26:18.570 --> 00:26:20.690
And that's emotional
intonation.

00:26:20.690 --> 00:26:23.620
And that brings us to
our reading for

00:26:23.620 --> 00:26:26.030
today, from Oliver Sacks.

00:26:26.030 --> 00:26:28.410
And I have to say, I think
Oliver Sacks is a phenomenal

00:26:28.410 --> 00:26:31.280
author, but there's always been
this sense of some of

00:26:31.280 --> 00:26:33.920
these stories are just a bit
too true to be-- just a bit

00:26:33.920 --> 00:26:34.870
cute, or whatever it is.

00:26:34.870 --> 00:26:39.920
And so for years, I taught
from this story, and I

00:26:39.920 --> 00:26:40.860
believed it's likely true.

00:26:40.860 --> 00:26:42.270
But I thought, well, it's
just a little cute.

00:26:42.270 --> 00:26:44.550
And then came an experiment
that showed it's not only

00:26:44.550 --> 00:26:46.650
cute, but it's exactly
correct.

00:26:46.650 --> 00:26:53.680
So for this, you appreciate this
a bit more if you know a

00:26:53.680 --> 00:26:55.220
couple things about Ronald
Reagan, who was

00:26:55.220 --> 00:26:56.400
president in the 1980s.

00:26:56.400 --> 00:27:00.380
So what do you guys know
about Ronald Reagan?

00:27:00.380 --> 00:27:01.702
What party?

00:27:01.702 --> 00:27:02.200
AUDIENCE: Republican.

00:27:02.200 --> 00:27:03.540
PROFESSOR: There you go.

00:27:03.540 --> 00:27:06.710
Was he considered a
good communicator?

00:27:06.710 --> 00:27:07.670
AUDIENCE: Yes.

00:27:07.670 --> 00:27:08.150
PROFESSOR: Yes.

00:27:08.150 --> 00:27:09.750
He was considered-- you know,
you can have all kinds of

00:27:09.750 --> 00:27:11.260
views on Ronald Reagan's
things.

00:27:11.260 --> 00:27:13.270
The public said, he's an
awesome communicator.

00:27:13.270 --> 00:27:15.322
In fact, Barack Obama said he
would like to communicate like

00:27:15.322 --> 00:27:16.880
Ronald Reagan.

00:27:16.880 --> 00:27:18.930
Whether that's to get Republican
votes or not,

00:27:18.930 --> 00:27:20.210
that's your judgment.

00:27:20.210 --> 00:27:22.520
But he was considered an
excellent communicator.

00:27:22.520 --> 00:27:27.280
Was he considered, by and large,
to be an extremely

00:27:27.280 --> 00:27:31.530
analytic thinker?

00:27:31.530 --> 00:27:32.680
He might have been.

00:27:32.680 --> 00:27:33.210
I don't know him.

00:27:33.210 --> 00:27:34.200
But not so much.

00:27:34.200 --> 00:27:36.225
Now, part of this is the
prejudice or not, but he was

00:27:36.225 --> 00:27:41.630
from California, and he was
an actor before he became

00:27:41.630 --> 00:27:43.200
president, or the governor
president.

00:27:43.200 --> 00:27:47.290
He also had some degree of--

00:27:47.290 --> 00:27:49.040
well, he ultimately got
Alzheimer's disease.

00:27:49.040 --> 00:27:50.750
It's a question about later in
his administration, whether he

00:27:50.750 --> 00:27:52.690
was already working his
way towards that.

00:27:52.690 --> 00:27:53.810
So anyway, complicated things.

00:27:53.810 --> 00:27:55.890
So he was a master communicator,
but not

00:27:55.890 --> 00:28:00.410
necessarily the Barack Obama,
Bill Clinton, Jimmy Carter

00:28:00.410 --> 00:28:04.450
version of an analytic power,
is the general perception.

00:28:04.450 --> 00:28:05.670
You never know whether
these are right or

00:28:05.670 --> 00:28:08.050
wrong for the person.

00:28:08.050 --> 00:28:10.840
But in the story, listening to
the president's speech, are

00:28:10.840 --> 00:28:13.220
patients with aphasia that
Oliver Sacks described, and

00:28:13.220 --> 00:28:16.460
he's getting laughter as they're
watching his speech.

00:28:16.460 --> 00:28:20.100
And what happens is, some of
the people with a left

00:28:20.100 --> 00:28:21.940
hemisphere lesion-- so you
have a language problem.

00:28:21.940 --> 00:28:24.980
We'll see a video of that
pretty shortly.

00:28:24.980 --> 00:28:30.740
They're laughing because they
think that the way he's

00:28:30.740 --> 00:28:31.600
talking and his facial

00:28:31.600 --> 00:28:34.030
expressions, they find hilarious.

00:28:34.030 --> 00:28:36.510
Now, why do they find
it hilarious?

00:28:36.510 --> 00:28:38.370
They have damage to the parts
of the brain that help them

00:28:38.370 --> 00:28:40.570
understand the content of
his message on the left

00:28:40.570 --> 00:28:41.450
hemisphere.

00:28:41.450 --> 00:28:44.260
And here's what Sacks
thinks is going on.

00:28:44.260 --> 00:28:46.370
"The feelings I have, which
all of us who work closely

00:28:46.370 --> 00:28:51.660
with aphasics have, is that you
cannot lie to an aphasic.

00:28:51.660 --> 00:28:56.050
He cannot grasp your words, and
so be deceived by them.

00:28:56.050 --> 00:28:58.820
But what he grasps, he grasps
with infallible precision,

00:28:58.820 --> 00:29:01.610
namely, the expression that
goes with the words, that

00:29:01.610 --> 00:29:04.920
total spontaneous, involuntary
expressiveness which can never

00:29:04.920 --> 00:29:09.250
be simulated or faked, or
words alone can hide all

00:29:09.250 --> 00:29:10.950
sometimes too easily."

00:29:10.950 --> 00:29:14.890
So when you give a very public
speech, especially under tough

00:29:14.890 --> 00:29:17.960
circumstances, that presidents
sometimes have to do, we

00:29:17.960 --> 00:29:20.160
sometimes think there's a
tension between exactly what

00:29:20.160 --> 00:29:23.260
they're saying and exactly
what they're thinking.

00:29:23.260 --> 00:29:25.560
And what they feel they need to
say, and exactly what they

00:29:25.560 --> 00:29:27.340
know is behind the scenes.

00:29:27.340 --> 00:29:31.070
And so the idea is that these
aphasics, they were saying

00:29:31.070 --> 00:29:33.570
this guy obviously doesn't
mean what he's saying.

00:29:33.570 --> 00:29:34.530
Look at his face.

00:29:34.530 --> 00:29:35.160
Look at his tone.

00:29:35.160 --> 00:29:38.780
I don't know what he's saying,
but it's hilarious, how he's

00:29:38.780 --> 00:29:41.800
masking his face and speaking
in a funny tone to tell

00:29:41.800 --> 00:29:44.510
something he truly
doesn't believe.

00:29:44.510 --> 00:29:45.860
That's their interpretation.

00:29:45.860 --> 00:29:47.850
And yet, he has an opposite
patient, a patient with a

00:29:47.850 --> 00:29:48.900
right hemisphere lesion.

00:29:48.900 --> 00:29:50.750
We said that's the side of
the hemisphere that seems

00:29:50.750 --> 00:29:53.920
important for facial expressions
and emotional

00:29:53.920 --> 00:29:54.780
intonation.

00:29:54.780 --> 00:29:58.520
And she complains about the
logic of the case he's making,

00:29:58.520 --> 00:29:59.910
as if having--

00:29:59.910 --> 00:30:01.430
normally, when we talk to
somebody, it's both.

00:30:01.430 --> 00:30:03.160
Left hemisphere's getting
the content,

00:30:03.160 --> 00:30:04.360
analyzing the content.

00:30:04.360 --> 00:30:06.450
Right hemisphere, something
about the intonation.

00:30:06.450 --> 00:30:07.855
True or not true?

00:30:07.855 --> 00:30:10.210
Are their eyes like this, and
they can't even face us as

00:30:10.210 --> 00:30:12.705
they tell us something
they know is false.

00:30:12.705 --> 00:30:14.920
But these patients, maybe
because they have one channel

00:30:14.920 --> 00:30:18.385
of information knocked out
because of injury, they become

00:30:18.385 --> 00:30:21.050
hypersensitive to
the other one.

00:30:21.050 --> 00:30:24.420
So I thought, well, that
sounds fantastically

00:30:24.420 --> 00:30:26.570
interesting, but is
that really true?

00:30:26.570 --> 00:30:28.870
So here in Boston, they did a
study with a large number of

00:30:28.870 --> 00:30:32.650
patients that says they had
superior identification of

00:30:32.650 --> 00:30:35.550
people who are lying to you--

00:30:35.550 --> 00:30:38.040
they had made video clips and
audio clips of people who are

00:30:38.040 --> 00:30:39.520
lying to you--

00:30:39.520 --> 00:30:42.400
if they had a left hemisphere
lesion.

00:30:42.400 --> 00:30:44.870
So this is a paper.

00:30:44.870 --> 00:30:46.180
So here's the two examples.

00:30:46.180 --> 00:30:47.880
So they have an example.

00:30:47.880 --> 00:30:49.350
Let's pick this one.

00:30:49.350 --> 00:30:51.090
Where both, you see
just a facial

00:30:51.090 --> 00:30:52.980
expression of a person talking.

00:30:52.980 --> 00:30:54.800
The voice is turned down.

00:30:54.800 --> 00:30:58.930
Your job is to decide whether
a person is lying or not.

00:30:58.930 --> 00:31:00.260
And some videos they're lying.

00:31:00.260 --> 00:31:01.470
Some videos they're not.

00:31:01.470 --> 00:31:03.640
So how would you do?

00:31:03.640 --> 00:31:05.840
I don't know how you would do,
but I can tell you how a group

00:31:05.840 --> 00:31:07.790
of Harvard undergraduates did.

00:31:07.790 --> 00:31:11.390
50% is chance, and here's
how they did.

00:31:11.390 --> 00:31:13.410
That was one of their
control groups.

00:31:13.410 --> 00:31:16.160
And, in fact, for studies like
this, they brought in many,

00:31:16.160 --> 00:31:21.310
many kinds of people to ask
whether anybody is better.

00:31:21.310 --> 00:31:24.240
Everybody's pretty much terrible
at directly telling

00:31:24.240 --> 00:31:25.270
whether a lie is going on.

00:31:25.270 --> 00:31:27.250
Now, these are video clips of
people you don't know in

00:31:27.250 --> 00:31:28.410
circumstances you don't know.

00:31:28.410 --> 00:31:29.370
That could matter.

00:31:29.370 --> 00:31:31.660
But everybody's terrible
at that.

00:31:31.660 --> 00:31:33.440
Policemen are terrible.

00:31:33.440 --> 00:31:35.150
FBI agents are terrible.

00:31:35.150 --> 00:31:37.720
The only group that seems to be
above chance, and not much

00:31:37.720 --> 00:31:40.640
above it, are Secret
Service agents.

00:31:40.640 --> 00:31:43.140
So if you're friend with a
Secret Service agent, and you

00:31:43.140 --> 00:31:45.740
want to deceive them, keep in
mind that they seem to be

00:31:45.740 --> 00:31:46.380
above chance.

00:31:46.380 --> 00:31:47.530
They're not awesome, but
they're above chance.

00:31:47.530 --> 00:31:50.030
But look who's doing better than
everybody, patients with

00:31:50.030 --> 00:31:52.380
the left hemisphere lesions.

00:31:52.380 --> 00:31:55.610
And the interpretation
is, all the content

00:31:55.610 --> 00:31:57.040
of the lie is gone.

00:31:57.040 --> 00:32:00.550
All they see is your expression,
or your expression

00:32:00.550 --> 00:32:01.940
and intonation.

00:32:01.940 --> 00:32:04.400
And because that's all they have
to go on, they can tell

00:32:04.400 --> 00:32:07.340
more when you have the
discomfort of lying than a

00:32:07.340 --> 00:32:10.800
person who's partly persuaded by
hearing the message you're

00:32:10.800 --> 00:32:12.380
sending as you talk.

00:32:12.380 --> 00:32:14.820
So I thought this was an overly
cute example, but it

00:32:14.820 --> 00:32:18.170
turns out to be exactly
replicable in a full study.

00:32:18.170 --> 00:32:19.420
Any questions about that?

00:32:21.970 --> 00:32:24.120
OK.

00:32:24.120 --> 00:32:25.690
So let's talk a little
bit about the

00:32:25.690 --> 00:32:26.840
brain basis of language.

00:32:26.840 --> 00:32:28.680
We've talked a little
bit about--

00:32:28.680 --> 00:32:31.910
we know that if you have damage
here, you have Broca's

00:32:31.910 --> 00:32:33.240
aphasia on the left hemisphere,

00:32:33.240 --> 00:32:35.390
trouble producing language.

00:32:35.390 --> 00:32:37.610
Damage here, Wernicke's
aphasia, trouble

00:32:37.610 --> 00:32:38.860
comprehending language.

00:32:42.010 --> 00:32:45.070
To a first approximation, and
there's a lot of research that

00:32:45.070 --> 00:32:47.960
has made this much more
complicated, but a first

00:32:47.960 --> 00:32:50.190
approximation, the lesions
are in frontal

00:32:50.190 --> 00:32:52.120
cortex or temporal cortex.

00:32:52.120 --> 00:32:54.730
The speech in these patients are
what's called non-fluent

00:32:54.730 --> 00:32:55.530
or telegraphic.

00:32:55.530 --> 00:32:57.690
They get single words out but
have a hard time with anything

00:32:57.690 --> 00:32:58.390
like a sentence.

00:32:58.390 --> 00:33:00.820
It can be very frustrating
for these patients.

00:33:00.820 --> 00:33:02.520
Wernicke's patients
seem much happier.

00:33:02.520 --> 00:33:03.650
You'll see a film in a moment.

00:33:03.650 --> 00:33:05.690
Their speech is fluent,
but it's

00:33:05.690 --> 00:33:09.100
amazingly empty of content.

00:33:09.100 --> 00:33:11.010
These patients have trouble
producing speech, but their

00:33:11.010 --> 00:33:12.720
comprehension is decent.

00:33:12.720 --> 00:33:15.310
These patients produce speech
pretty well, but their

00:33:15.310 --> 00:33:16.050
comprehension.

00:33:16.050 --> 00:33:20.320
So they're mirror opposites.

00:33:20.320 --> 00:33:21.400
And let's see, what do I want?

00:33:21.400 --> 00:33:21.970
Yes.

00:33:21.970 --> 00:33:26.620
So in this field, a frequently,
incredibly

00:33:26.620 --> 00:33:29.240
frequently used picture
to study aphasia is

00:33:29.240 --> 00:33:30.710
this specific picture.

00:33:30.710 --> 00:33:32.610
It's called the cookie
theft picture.

00:33:32.610 --> 00:33:35.660
Why it's become widely used,
I'm not exactly sure.

00:33:35.660 --> 00:33:38.920
But once it's widely used, then
often everybody uses it.

00:33:38.920 --> 00:33:41.620
So they'll put a picture like
this in front of an

00:33:41.620 --> 00:33:43.940
individual, and they'll
ask them to describe

00:33:43.940 --> 00:33:44.530
what's going on.

00:33:44.530 --> 00:33:46.710
It's to elicit production.

00:33:46.710 --> 00:33:50.275
And you would say something
like, there's two kids.

00:33:50.275 --> 00:33:51.860
The kid's on a stool.

00:33:51.860 --> 00:33:53.020
Uh oh, is he going
to fall over?

00:33:53.020 --> 00:33:54.830
Because they're falling off
the cookie jar, and the

00:33:54.830 --> 00:33:56.280
mother's sort of not
paying attention.

00:33:56.280 --> 00:33:58.370
Her mind must be somewhere else
because she's letting the

00:33:58.370 --> 00:34:00.310
water splish splash
here, right?

00:34:00.310 --> 00:34:02.810
You would say something
like that.

00:34:02.810 --> 00:34:05.150
So let's see what patients with
aphasia say, and other

00:34:05.150 --> 00:34:06.960
examples of their difficulty
with language.

00:34:16.400 --> 00:34:17.674
Yeah?

00:34:17.674 --> 00:34:20.290
AUDIENCE: [INAUDIBLE].

00:34:20.290 --> 00:34:23.100
PROFESSOR: Their writing tends
to go pretty much with their

00:34:23.100 --> 00:34:24.840
spoken language.

00:34:24.840 --> 00:34:27.310
That's a really good question
about-- so the question is,

00:34:27.310 --> 00:34:28.699
what would their writing
be like?

00:34:28.699 --> 00:34:33.952
It very much parallels their
spoken language.

00:34:33.952 --> 00:34:34.430
Yeah?

00:34:34.430 --> 00:34:37.256
AUDIENCE: [INAUDIBLE].

00:34:37.256 --> 00:34:38.130
PROFESSOR: Sorry?

00:34:38.130 --> 00:34:38.990
AUDIENCE: [INAUDIBLE]

00:34:38.990 --> 00:34:40.280
sign language?

00:34:40.280 --> 00:34:42.830
PROFESSOR: OK, so very
good question.

00:34:42.830 --> 00:34:46.030
Could not fluent aphasics learn
sign language to bypass,

00:34:46.030 --> 00:34:47.780
in a sense, their
problem, right?

00:34:47.780 --> 00:34:50.020
So I'm about to answer
that in a moment,

00:34:50.020 --> 00:34:51.360
right on the next slide.

00:34:51.360 --> 00:34:54.710
Not directly that, but the
implication of that.

00:34:54.710 --> 00:34:56.610
Because the important thing is--
so let me show you this

00:34:56.610 --> 00:34:57.220
thing, and let me come back.

00:34:57.220 --> 00:34:59.650
The question is, could you train
an aphasic to use sign

00:34:59.650 --> 00:35:02.700
language to get around his or
her language problem, if it's

00:35:02.700 --> 00:35:03.770
a speech problem?

00:35:03.770 --> 00:35:10.180
So people have studied sign
language as a language for

00:35:10.180 --> 00:35:12.870
many reasons, but a
huge one is this.

00:35:12.870 --> 00:35:15.560
Because I'm going to interpret
your question the way I need

00:35:15.560 --> 00:35:18.000
to, and then you can tell me
if I did it right from your

00:35:18.000 --> 00:35:18.310
perspective.

00:35:18.310 --> 00:35:23.530
Which is, in what sense does
language equal speech?

00:35:23.530 --> 00:35:27.010
Could you communicate
linguistically like, but just

00:35:27.010 --> 00:35:30.300
switch to your hands, if for
some reason-- and the answer

00:35:30.300 --> 00:35:33.290
is, that at the level of the
problem of these patients,

00:35:33.290 --> 00:35:34.880
they could not do it.

00:35:34.880 --> 00:35:37.800
Because these kinds of patients,
the problem isn't

00:35:37.800 --> 00:35:38.810
really speech.

00:35:38.810 --> 00:35:41.600
There are people who have
problems with the motor

00:35:41.600 --> 00:35:42.730
apparatus to move their mouth.

00:35:42.730 --> 00:35:44.900
That does exist.

00:35:44.900 --> 00:35:48.040
Their problem is in language
processes that produce your

00:35:48.040 --> 00:35:53.200
speech intent, or language
processes that extract meaning

00:35:53.200 --> 00:35:54.310
from sounds.

00:35:54.310 --> 00:35:57.260
And those language processes
that produce what you want to

00:35:57.260 --> 00:36:01.440
say, that comprehend what you
hear or read, those are the

00:36:01.440 --> 00:36:04.420
same ones that are utilized,
as far as we understand,

00:36:04.420 --> 00:36:06.870
whether you're producing
language in the typical, vocal

00:36:06.870 --> 00:36:09.050
way, or in the less
typical, but

00:36:09.050 --> 00:36:12.130
well-studied, way of sign language.

00:36:12.130 --> 00:36:12.530
Does that make sense?

00:36:12.530 --> 00:36:13.010
OK.

00:36:13.010 --> 00:36:14.170
So you can't get around
it because

00:36:14.170 --> 00:36:16.440
it's a language problem.

00:36:16.440 --> 00:36:19.550
So both by brain imaging
evidence, for example, here

00:36:19.550 --> 00:36:22.660
are patients who use sign
language turning on Broca's

00:36:22.660 --> 00:36:25.490
area and Wernicke's area
on tasks that are about

00:36:25.490 --> 00:36:26.575
comprehension and production.

00:36:26.575 --> 00:36:29.550
It's the same areas, whether
it's using your hands for sign

00:36:29.550 --> 00:36:33.840
language or using your mouth to
talk or your ear to hear.

00:36:33.840 --> 00:36:36.070
Further, there have been really
fun studies where they

00:36:36.070 --> 00:36:39.960
took deaf infants and have
shown-- and it's not easy to

00:36:39.960 --> 00:36:42.750
show, but I think most people
are convinced that they babble

00:36:42.750 --> 00:36:43.370
with their hands.

00:36:43.370 --> 00:36:45.530
The same way babies make
little sounds to get

00:36:45.530 --> 00:36:46.260
themselves going.

00:36:46.260 --> 00:36:48.720
We'll talk more about
infants in a moment.

00:36:48.720 --> 00:36:51.030
Infants who grow up in an
environment where they only

00:36:51.030 --> 00:36:55.430
learn sign language babble
with their hands because

00:36:55.430 --> 00:36:58.880
speech, language is not about
your mouth talking.

00:36:58.880 --> 00:37:02.420
It's about ideas that you can
communicate by your production

00:37:02.420 --> 00:37:04.540
or comprehension, by your
hand or by your mouth.

00:37:04.540 --> 00:37:08.860
So language is separable from
speech, because even without

00:37:08.860 --> 00:37:12.010
speech, you communicate in very
much the same way, in

00:37:12.010 --> 00:37:14.470
terms of sign language,
by many studies.

00:37:14.470 --> 00:37:15.060
Yeah?

00:37:15.060 --> 00:37:19.326
AUDIENCE: So does that mean that
people that [INAUDIBLE]

00:37:19.326 --> 00:37:22.650
imagined words or language?

00:37:22.650 --> 00:37:23.405
PROFESSOR: Say that
again, sorry?

00:37:23.405 --> 00:37:25.283
AUDIENCE: Does that mean that
they also couldn't imagine

00:37:25.283 --> 00:37:26.209
words or language?

00:37:26.209 --> 00:37:27.135
PROFESSOR: Who couldn't?

00:37:27.135 --> 00:37:29.510
AUDIENCE: People with Broca's
area aphasia.

00:37:33.190 --> 00:37:35.010
PROFESSOR: I don't think so.

00:37:35.010 --> 00:37:36.720
We talked about this
before, that--

00:37:36.720 --> 00:37:39.820
I may not be understanding your
question exactly, but--

00:37:39.820 --> 00:37:42.890
in many ways, imagining
something is the same part of

00:37:42.890 --> 00:37:43.890
the brain that does it.

00:37:43.890 --> 00:37:45.800
When we imagine a face, we
turn on the parts of

00:37:45.800 --> 00:37:46.720
our brain that sees.

00:37:46.720 --> 00:37:49.300
I think when you imagine,
if you--

00:37:49.300 --> 00:37:50.300
I do this sometimes.

00:37:50.300 --> 00:37:51.510
I say, here's what I'm
going to say to that

00:37:51.510 --> 00:37:53.220
person, in my head.

00:37:53.220 --> 00:37:54.770
And I practice that
little speech.

00:37:54.770 --> 00:37:58.740
That's this part of the brain
pretty much running.

00:37:58.740 --> 00:38:01.800
There's just a last little step
of moving your mouth, but

00:38:01.800 --> 00:38:03.270
from the discussion we're having
now, that's kind of a

00:38:03.270 --> 00:38:06.050
superficial thing of
moving your mouth.

00:38:06.050 --> 00:38:09.170
That's not speech planning,
speech thinking, speech

00:38:09.170 --> 00:38:10.110
intentions.

00:38:10.110 --> 00:38:14.130
Those are our core process for
production, for comprehension,

00:38:14.130 --> 00:38:17.870
and I think imagined stuff
is exactly the same way.

00:38:17.870 --> 00:38:18.650
Is that OK?

00:38:18.650 --> 00:38:19.100
Yeah.

00:38:19.100 --> 00:38:20.093
Yeah?

00:38:20.093 --> 00:38:22.804
AUDIENCE: So if they can't
really imagine language, then

00:38:22.804 --> 00:38:25.516
how do they-- what is
their thinking like?

00:38:25.516 --> 00:38:26.766
[INAUDIBLE]?

00:38:30.460 --> 00:38:35.030
PROFESSOR: Yeah, that's a good--
so the question is, if

00:38:35.030 --> 00:38:37.820
they have these troubles with
core language ideas,

00:38:37.820 --> 00:38:39.870
abilities, what's their
thinking like?

00:38:39.870 --> 00:38:43.660
I could just tell you, it's a
classic debate for humans, how

00:38:43.660 --> 00:38:45.580
deeply language equals
thought.

00:38:45.580 --> 00:38:48.420
But language [INAUDIBLE]
in a sense of speech.

00:38:48.420 --> 00:38:50.790
It's just one of those things
that people debate a lot

00:38:50.790 --> 00:38:53.000
because it depends what you mean
by these things in many

00:38:53.000 --> 00:38:55.360
different ways.

00:38:55.360 --> 00:38:57.490
But if language means something
like comprehending

00:38:57.490 --> 00:39:00.290
ideas in the world, it's hard
to get around that for

00:39:00.290 --> 00:39:01.560
thinking [INAUDIBLE].

00:39:01.560 --> 00:39:04.360
It means planning actions in
the world that change the

00:39:04.360 --> 00:39:08.190
world around you, and
communicating that, it's hard

00:39:08.190 --> 00:39:11.410
to not have that capacity for
thinking in that way, right?

00:39:11.410 --> 00:39:14.900
So it's at this level, not
whether you use your hands or

00:39:14.900 --> 00:39:17.990
your mouth, it's very hard to
separate out what you might

00:39:17.990 --> 00:39:21.310
call thought and what you
might call language.

00:39:21.310 --> 00:39:23.550
But keep in mind that these
patients that we see, it's not

00:39:23.550 --> 00:39:26.040
like they lost all of
language at all.

00:39:26.040 --> 00:39:28.100
So even the Broca's aphasic
has trouble producing

00:39:28.100 --> 00:39:30.040
language, he's producing
some, and he

00:39:30.040 --> 00:39:31.420
comprehends it pretty well.

00:39:31.420 --> 00:39:33.350
So it's not like these patients
have lost all of

00:39:33.350 --> 00:39:35.840
their language circuitry, and
therefore don't have thought

00:39:35.840 --> 00:39:36.820
available, right?

00:39:36.820 --> 00:39:39.260
They have a lot of language left
in their head, and a lot

00:39:39.260 --> 00:39:40.460
of thought left in their mind.

00:39:40.460 --> 00:39:43.750
Is that OK?

00:39:43.750 --> 00:39:45.865
So one interesting thing that
people have wondered, and

00:39:45.865 --> 00:39:47.770
where brain imaging gives us
evidence that we couldn't have

00:39:47.770 --> 00:39:50.930
imagined 15 years
ago, is this.

00:39:50.930 --> 00:39:53.180
So if you have damage, let's
say, here's a patient with

00:39:53.180 --> 00:39:56.290
damage to the left
frontal area.

00:39:56.290 --> 00:39:59.550
And they get better because
they've had a stroke here or

00:39:59.550 --> 00:40:00.190
an injury here.

00:40:00.190 --> 00:40:01.170
They get better.

00:40:01.170 --> 00:40:02.850
A question that people have
wondered is, do they get

00:40:02.850 --> 00:40:05.860
better by sort of somehow
recovering ability in a part

00:40:05.860 --> 00:40:07.470
of their brain that
got injured?

00:40:07.470 --> 00:40:10.120
Or do they get better because
they use an entirely different

00:40:10.120 --> 00:40:11.820
pathway in their brain?

00:40:11.820 --> 00:40:13.340
Now, until brain imaging,
you couldn't begin

00:40:13.340 --> 00:40:14.490
to touch that question.

00:40:14.490 --> 00:40:17.220
But now we have some evidence,
and I'll make it just--

00:40:17.220 --> 00:40:21.750
and so here's an example of a
patient who has damage here.

00:40:21.750 --> 00:40:25.190
They have him do a task that
in healthy people turns on

00:40:25.190 --> 00:40:28.330
exactly that place.

00:40:28.330 --> 00:40:30.780
And when he does it,
he can do it OK.

00:40:30.780 --> 00:40:31.950
And where does he do it?

00:40:31.950 --> 00:40:33.630
In his right hemisphere.

00:40:33.630 --> 00:40:36.865
So this patient seems to switch
which hemisphere he

00:40:36.865 --> 00:40:37.730
uses for language.

00:40:37.730 --> 00:40:40.650
It takes time and training
to recover that.

00:40:40.650 --> 00:40:44.070
But it seems like he's found
an alternate pathway to

00:40:44.070 --> 00:40:46.900
produce a language somehow.

00:40:46.900 --> 00:40:49.920
Turns out that the more people
have studied these things, the

00:40:49.920 --> 00:40:51.980
more complicated it gets.

00:40:51.980 --> 00:40:54.260
Maybe just diversity of people,
diversity of injuries,

00:40:54.260 --> 00:40:55.140
or whatever.

00:40:55.140 --> 00:40:57.030
So it doesn't always
work this way.

00:40:57.030 --> 00:40:59.350
But often, it seems like when
people have better recovery

00:40:59.350 --> 00:41:02.990
from a large injury to the
speaking left hemisphere, they

00:41:02.990 --> 00:41:06.480
find a way, somehow, through
practice and training, to get

00:41:06.480 --> 00:41:08.100
the right hemisphere to take
over some of those

00:41:08.100 --> 00:41:09.350
responsibilities.

00:41:12.080 --> 00:41:13.010
Language acquisition.

00:41:13.010 --> 00:41:15.920
So one of the most amazing
things about infants is they

00:41:15.920 --> 00:41:20.600
go from zero to speech in just
a couple of years, and go

00:41:20.600 --> 00:41:22.910
through an unbelievable training
and program of

00:41:22.910 --> 00:41:24.120
development.

00:41:24.120 --> 00:41:25.410
So what happens?

00:41:25.410 --> 00:41:28.060
In two or three months, they
perceive all the phonemes.

00:41:28.060 --> 00:41:30.130
They notice this change
in phonemes.

00:41:30.130 --> 00:41:32.510
By six months, they start to
ignore distinctions among

00:41:32.510 --> 00:41:34.130
sounds that are not used
in their language.

00:41:34.130 --> 00:41:36.680
In other words, other
language sounds.

00:41:36.680 --> 00:41:39.590
They start to babble, in
some way getting their

00:41:39.590 --> 00:41:41.050
speech ready to go.

00:41:41.050 --> 00:41:43.520
By eight months, they start to
identify single words in a

00:41:43.520 --> 00:41:44.570
speech stream.

00:41:44.570 --> 00:41:46.500
In a year, their babbling
becomes

00:41:46.500 --> 00:41:48.070
more adult-like sounding.

00:41:48.070 --> 00:41:51.880
But a little bit after that,
they get up to about 50 words

00:41:51.880 --> 00:41:53.070
in understanding.

00:41:53.070 --> 00:41:56.780
So comprehension always goes
ahead of production.

00:41:56.780 --> 00:41:58.920
Children understand words before
they can speak them.

00:41:58.920 --> 00:41:59.970
But they're lagging.

00:41:59.970 --> 00:42:02.030
Here comes a speech, telegraphic
speech, short

00:42:02.030 --> 00:42:03.610
sentences at two years.

00:42:03.610 --> 00:42:06.330
And then, finally, pretty
complete speech, in

00:42:06.330 --> 00:42:07.740
many ways, by nine.

00:42:07.740 --> 00:42:12.570
So it's a long, long process,
learning 10,000 words

00:42:12.570 --> 00:42:14.010
estimated by the age of six.

00:42:14.010 --> 00:42:16.470
That's a spectacular
learning curve.

00:42:16.470 --> 00:42:18.590
When they do experiments on
infants where they measure

00:42:18.590 --> 00:42:20.980
things like preference or
learning by sucking rates on a

00:42:20.980 --> 00:42:23.030
nipple because they can't talk,
they try to figure out,

00:42:23.030 --> 00:42:24.430
what is the infant hearing.

00:42:24.430 --> 00:42:27.920
Within two hours of birth, they
suck with more enthusiasm

00:42:27.920 --> 00:42:30.200
to their mother's voice
over another voice.

00:42:30.200 --> 00:42:31.410
These are infants, this
is their first

00:42:31.410 --> 00:42:32.930
visit to their mother.

00:42:32.930 --> 00:42:36.330
You have to sign up to
do this experiment.

00:42:36.330 --> 00:42:38.040
Now, they might have heard
their mother's voice in a

00:42:38.040 --> 00:42:44.200
certain way in the
womb, right?

00:42:44.200 --> 00:42:46.500
But they already have picked
up, by three-day-old, they

00:42:46.500 --> 00:42:49.050
prefer language to many other
sounds, like music.

00:42:49.050 --> 00:42:51.240
We come to love music, but
that's not the urgent business

00:42:51.240 --> 00:42:53.060
of the infant.

00:42:53.060 --> 00:42:55.780
By four days old, they start
to notice the differences

00:42:55.780 --> 00:42:56.930
between languages.

00:42:56.930 --> 00:42:59.180
This is amazing.

00:42:59.180 --> 00:43:01.050
Their rate of learning
is fantastic.

00:43:01.050 --> 00:43:03.810
And by two months, they can tell
the ba-ga distinction,

00:43:03.810 --> 00:43:07.350
subtle distinctions, 40
milliseconds information.

00:43:07.350 --> 00:43:09.970
And then they develop a
continuingly strong preference

00:43:09.970 --> 00:43:12.710
for the sounds of their
native language.

00:43:12.710 --> 00:43:15.760
Now, some years ago, people
debated endlessly this left

00:43:15.760 --> 00:43:17.800
hemisphere specialization
for language.

00:43:17.800 --> 00:43:19.320
This is fantastically essential

00:43:19.320 --> 00:43:20.990
and striking in humans.

00:43:20.990 --> 00:43:24.630
Does that grow over time,
or is it in your mind

00:43:24.630 --> 00:43:26.400
and brain at birth?

00:43:26.400 --> 00:43:29.080
And the evidence is pretty
compelling for many sources,

00:43:29.080 --> 00:43:30.430
but here is maybe the most.

00:43:30.430 --> 00:43:33.680
Here's fMRI on two-day-old
infants.

00:43:33.680 --> 00:43:36.630
You might ask, how do they
instruct the infants to lie

00:43:36.630 --> 00:43:39.110
still in the scanner and
participate in the experiment?

00:43:39.110 --> 00:43:43.110
And the answer is, at two days
old, infants mostly just lay

00:43:43.110 --> 00:43:44.410
there and don't do much.

00:43:47.750 --> 00:43:48.390
They don't do much.

00:43:48.390 --> 00:43:49.590
And there's a lot going on in
their mind, but they don't

00:43:49.590 --> 00:43:51.390
have much physical apparatus
to do much.

00:43:51.390 --> 00:43:54.870
So you could swaddle them very
warmly, and just lay--

00:43:54.870 --> 00:43:57.050
you put them on a cradle
into the fMRI scanner.

00:43:57.050 --> 00:43:59.970
Their parents have signed
up for them.

00:43:59.970 --> 00:44:02.760
And you could play them sounds
while they just lay there and

00:44:02.760 --> 00:44:03.240
do nothing.

00:44:03.240 --> 00:44:04.180
They can't push buttons.

00:44:04.180 --> 00:44:04.980
They can't do stuff.

00:44:04.980 --> 00:44:06.640
But you can see what's happening
inside their brain.

00:44:06.640 --> 00:44:09.520
And at two days old, here's the
left hemisphere response

00:44:09.520 --> 00:44:11.220
to language compared
to the right.

00:44:11.220 --> 00:44:14.870
So it's there instantly,
we believe.

00:44:14.870 --> 00:44:17.740
It's been there, actually in the
womb as the child's been

00:44:17.740 --> 00:44:18.600
developing.

00:44:18.600 --> 00:44:21.470
This left hemisphere,
genetically set up dominance

00:44:21.470 --> 00:44:22.840
for language in almost
everybody.

00:44:26.810 --> 00:44:29.330
fMRI, just for those
of you who--

00:44:29.330 --> 00:44:31.660
again, we can start doing it
at about age five, but from

00:44:31.660 --> 00:44:34.720
about three months to age five,
it's incredibly hard to

00:44:34.720 --> 00:44:36.530
do children in a scanner.

00:44:36.530 --> 00:44:39.410
But the first couple days,
it's not so hard.

00:44:39.410 --> 00:44:41.730
Now, a really interesting
line of research has

00:44:41.730 --> 00:44:42.750
talked about motherese.

00:44:42.750 --> 00:44:43.620
Have you heard this term?

00:44:43.620 --> 00:44:46.210
So how do grown-ups,
even college

00:44:46.210 --> 00:44:47.720
students, talk to babies?

00:44:47.720 --> 00:44:49.215
Do you talk to babies
just like you talk

00:44:49.215 --> 00:44:50.600
to everybody else?

00:44:50.600 --> 00:44:51.850
No.

00:44:53.790 --> 00:44:55.710
Cute baby.

00:44:55.710 --> 00:44:56.960
Baby, baby.

00:44:56.960 --> 00:44:58.210
And everybody looks like
an idiot, right?

00:45:00.740 --> 00:45:02.850
And you could just think, OK,
that's just people being

00:45:02.850 --> 00:45:03.900
idiots around babies.

00:45:03.900 --> 00:45:06.690
But it turns out there's been
beautiful, beautiful studies

00:45:06.690 --> 00:45:10.290
to show that, while you
spontaneously do this-- and

00:45:10.290 --> 00:45:14.010
you and I don't take motherese
or fatherese courses--

00:45:14.010 --> 00:45:18.070
but what we do spontaneously
is that way of elongating

00:45:18.070 --> 00:45:22.520
sounds, the intonations we give,
are incredibly perfect

00:45:22.520 --> 00:45:24.340
to help the baby
learn language.

00:45:24.340 --> 00:45:27.710
We're exactly doing what
they can understand.

00:45:27.710 --> 00:45:29.770
And that, again, must
be virtually in

00:45:29.770 --> 00:45:31.610
our genes, I think.

00:45:31.610 --> 00:45:33.540
We hear it around,
so it's not zero.

00:45:33.540 --> 00:45:37.660
But the short pauses, careful
enunciation, exaggerated

00:45:37.660 --> 00:45:39.930
intonation, that's--

00:45:39.930 --> 00:45:41.880
you could do experiments where
you take language and you get

00:45:41.880 --> 00:45:43.380
rid of those, and the babies
are like-- they don't know

00:45:43.380 --> 00:45:43.860
what's going on.

00:45:43.860 --> 00:45:44.770
They're not interested.

00:45:44.770 --> 00:45:46.710
You put them back in, boom.

00:45:46.710 --> 00:45:48.950
Those are perfect
for the baby.

00:45:48.950 --> 00:45:52.610
It's amazing, not only how
babies are ready to go, but

00:45:52.610 --> 00:45:55.090
how adults who talk with
them are ready to go.

00:45:55.090 --> 00:45:58.210
I mean, it's just amazing.

00:45:58.210 --> 00:46:01.240
None of us will ever be as good
a teacher in a classroom

00:46:01.240 --> 00:46:03.690
as all of you will be
with a baby anytime

00:46:03.690 --> 00:46:04.490
you talk with them.

00:46:04.490 --> 00:46:06.500
It's just amazing.

00:46:06.500 --> 00:46:08.260
So there's beautiful experiments
that show that

00:46:08.260 --> 00:46:11.950
every time you tinker with the
way the mother spontaneously,

00:46:11.950 --> 00:46:14.930
or parents spontaneously, talk
to indants, you make it worse,

00:46:14.930 --> 00:46:16.750
less interesting, and less
informative for the infant.

00:46:19.400 --> 00:46:22.660
So now let's talk about a
more formal experiment.

00:46:22.660 --> 00:46:25.030
These are the classic experiment
that nailed home

00:46:25.030 --> 00:46:28.710
this message, that at birth
we're universal understanders

00:46:28.710 --> 00:46:30.260
of all sounds in the world.

00:46:30.260 --> 00:46:33.800
And as the years go along, we
became only good at the sounds

00:46:33.800 --> 00:46:37.220
that are relevant, that are
part of our own language.

00:46:37.220 --> 00:46:41.380
A famous example in the US is
a difficulty for Japanese

00:46:41.380 --> 00:46:43.600
native speakers for the
R-L distinction.

00:46:43.600 --> 00:46:46.660
But every language has some
distinctions it doesn't make,

00:46:46.660 --> 00:46:48.660
and they're really hard
when you get older.

00:46:48.660 --> 00:46:49.780
So here's the idea.

00:46:49.780 --> 00:46:53.450
In English, we make a
distinction between ba and da,

00:46:53.450 --> 00:46:55.800
and by 10 to 12 months, an
infant can make that

00:46:55.800 --> 00:46:56.660
distinction.

00:46:56.660 --> 00:46:57.990
And certainly, adults can.

00:46:57.990 --> 00:47:01.330
That's categorical perception
that we talked about, that

00:47:01.330 --> 00:47:04.990
kind of division between
ba and da.

00:47:04.990 --> 00:47:07.530
But there's sounds that occur
in other language.

00:47:07.530 --> 00:47:09.640
So we're shown here.

00:47:09.640 --> 00:47:10.680
And you're going to hear--

00:47:10.680 --> 00:47:13.060
some of you may know Hindi
sounds, and it'll be a

00:47:13.060 --> 00:47:14.850
different experience for you.

00:47:14.850 --> 00:47:17.760
Every year, when I first hear
this clip I'm about to show

00:47:17.760 --> 00:47:19.910
you, I think something's
wrong.

00:47:19.910 --> 00:47:21.640
Because they say, the infant's
going to tell the difference

00:47:21.640 --> 00:47:23.240
between da and da.

00:47:23.240 --> 00:47:24.500
And I go, well, how are
they going to do that?

00:47:24.500 --> 00:47:25.880
I can't do that.

00:47:25.880 --> 00:47:28.530
That's because they really are
different sounds you're

00:47:28.530 --> 00:47:31.570
hearing, but in English, they're
not different sounds.

00:47:31.570 --> 00:47:33.730
And so it all sounds
the same to me.

00:47:33.730 --> 00:47:35.540
I'm no longer a universal
learner.

00:47:35.540 --> 00:47:38.620
I haven't been since
quite a while.

00:47:38.620 --> 00:47:39.870
So here we go.

00:47:43.750 --> 00:47:46.370
I'm stunned, always, by the
brilliance of infant

00:47:46.370 --> 00:47:50.080
researchers like this, who
somehow, pretty convincingly

00:47:50.080 --> 00:47:53.770
in most cases, really tell you
what the infant knows or pays

00:47:53.770 --> 00:47:56.103
attention to or thinks in pretty
compelling ways, even

00:47:56.103 --> 00:47:59.290
though they can't tell
you themselves.

00:47:59.290 --> 00:48:01.310
So one thing that's come
up in the US--

00:48:01.310 --> 00:48:04.215
I mean, the whole world is
becoming ever more bilingual

00:48:04.215 --> 00:48:06.570
and trilingual.

00:48:06.570 --> 00:48:08.473
There used to be some concerns,
less now, and I'll

00:48:08.473 --> 00:48:10.940
tell you, you don't have
to worry about it.

00:48:10.940 --> 00:48:13.200
Whether it's bad to learn
two languages at once

00:48:13.200 --> 00:48:13.870
when you're a kid.

00:48:13.870 --> 00:48:16.016
And some parents worry
about this, right?

00:48:16.016 --> 00:48:17.510
You're going to be all messed
up because you're going to

00:48:17.510 --> 00:48:19.720
blend them all up and not
do well in school.

00:48:19.720 --> 00:48:22.345
And so there's been quite a
bit of research, actually,

00:48:22.345 --> 00:48:25.270
tracking and asking whether
there's any downside.

00:48:25.270 --> 00:48:26.590
I mean, there's huge
upsides to knowing

00:48:26.590 --> 00:48:27.860
two or three languages.

00:48:27.860 --> 00:48:30.980
Are there any downsides in terms
of language development?

00:48:30.980 --> 00:48:35.540
And the objective evidence is
pretty compelling that for

00:48:35.540 --> 00:48:40.300
babbling, for first words, for
pairs of words at a time,

00:48:40.300 --> 00:48:42.680
[INAUDIBLE] at two months, that
there's really no cost.

00:48:42.680 --> 00:48:45.660
So they followed, in Canada, for
example, children who grew

00:48:45.660 --> 00:48:48.830
up learning English and French,
English and ASL,

00:48:48.830 --> 00:48:51.800
French or English only.

00:48:51.800 --> 00:48:54.270
Everybody does the same,
pretty much.

00:48:54.270 --> 00:48:56.950
There's no cost, compared
to the huge gain.

00:48:56.950 --> 00:48:59.055
Sometimes some kids will
get a little mixed up.

00:48:59.055 --> 00:49:01.470
They'll actually mix up
languages because they'll just

00:49:01.470 --> 00:49:03.860
mix them up when they're
pretty young.

00:49:03.860 --> 00:49:06.560
But there's no developmental
delay of milestones or

00:49:06.560 --> 00:49:07.700
negative consequences.

00:49:07.700 --> 00:49:10.920
There's only the benefit of
having two or three languages

00:49:10.920 --> 00:49:11.690
that you could speak.

00:49:11.690 --> 00:49:13.430
So the evidence is overwhelming,
empirically,

00:49:13.430 --> 00:49:15.920
that there's no cost for
learning more than one

00:49:15.920 --> 00:49:19.790
language at a time when
you're an infant.

00:49:19.790 --> 00:49:23.070
One thing, there is a cost, a
little bit, if you're a boy.

00:49:23.070 --> 00:49:25.020
So unknown why this happened.

00:49:25.020 --> 00:49:27.400
I don't think there's any
deep science about it.

00:49:27.400 --> 00:49:30.320
But everybody has observed that,
for example, number of

00:49:30.320 --> 00:49:32.560
new words understood during the
first two years of life,

00:49:32.560 --> 00:49:35.100
that girls, on average,
outperform boys in language

00:49:35.100 --> 00:49:36.350
acquisition early on.

00:49:36.350 --> 00:49:38.600
And then things get closer,
pretty much.

00:49:38.600 --> 00:49:40.500
But the rate of learning
on girls on average

00:49:40.500 --> 00:49:42.280
is faster than boys.

00:49:42.280 --> 00:49:46.570
Nobody, I think, understands
the specific basis of that.

00:49:46.570 --> 00:49:48.970
So when people talk about
critical periods in language

00:49:48.970 --> 00:49:52.930
acquisition, and what are
fragile or resilient aspects

00:49:52.930 --> 00:49:53.530
of language.

00:49:53.530 --> 00:49:56.320
Which aspects of learning
language that seem very

00:49:56.320 --> 00:49:58.910
dependent on the age you are,
and which are independent.

00:49:58.910 --> 00:50:02.380
So phonology, including the
production of language sounds,

00:50:02.380 --> 00:50:03.490
seems very fragile.

00:50:03.490 --> 00:50:05.510
Easy when you're young, harder
when you're older.

00:50:05.510 --> 00:50:07.990
Grammar is also age-sensitive,
making subtle grammatical

00:50:07.990 --> 00:50:10.080
distinctions.

00:50:10.080 --> 00:50:13.180
Learning about meaning or
vocabulary, you're perfectly

00:50:13.180 --> 00:50:14.140
fine when you're an adult.

00:50:14.140 --> 00:50:15.800
That doesn't seem to
show something

00:50:15.800 --> 00:50:17.290
like a sensitive period.

00:50:17.290 --> 00:50:19.390
And here's one famous study
that established

00:50:19.390 --> 00:50:19.980
some of these things.

00:50:19.980 --> 00:50:23.850
They looked at what age people
arrive to different countries,

00:50:23.850 --> 00:50:25.260
from different countries
to the US and

00:50:25.260 --> 00:50:26.300
started learning English.

00:50:26.300 --> 00:50:29.650
These are people that didn't
know English before.

00:50:29.650 --> 00:50:33.440
They're all adults now, and
they had them do quickly

00:50:33.440 --> 00:50:35.120
grammatical judgments.

00:50:35.120 --> 00:50:37.860
And what they found is, if you
came to the US when you--

00:50:37.860 --> 00:50:40.560
if you were born in the US, or
you came to the age three to

00:50:40.560 --> 00:50:43.650
seven, you were very
good at doing this.

00:50:43.650 --> 00:50:47.590
But the later you came here, 8
to 10 years of age, 11 to 15,

00:50:47.590 --> 00:50:52.270
17 to 39, your ability to make
these fast, subtle grammatical

00:50:52.270 --> 00:50:54.320
judgments just went down.

00:50:54.320 --> 00:50:56.165
Even though you've been here, in
many cases, for many years.

00:50:56.165 --> 00:50:57.450
You're very smart.

00:50:57.450 --> 00:50:58.770
Accents are like that, right?

00:50:58.770 --> 00:51:00.990
People come to the US at a
certain point in their

00:51:00.990 --> 00:51:01.960
development.

00:51:01.960 --> 00:51:03.580
They're very skilled in
English, have a big

00:51:03.580 --> 00:51:06.170
vocabulary, but the accent
never disappears.

00:51:06.170 --> 00:51:08.800
That's another thing in terms
of language production that

00:51:08.800 --> 00:51:12.250
has its fragile developmental
early period.

00:51:12.250 --> 00:51:13.810
So the last thing I'll do, and
then show you a movie for

00:51:13.810 --> 00:51:17.530
about eight minutes.

00:51:17.530 --> 00:51:20.190
We get this occasional
horrible experiments.

00:51:20.190 --> 00:51:21.650
It's not even an experiment
of nature.

00:51:21.650 --> 00:51:25.120
It's just an experiment of
horrible human behavior.

00:51:25.120 --> 00:51:27.385
And the most famous one in
language development, which

00:51:27.385 --> 00:51:28.660
had a huge impact on
the field at the

00:51:28.660 --> 00:51:30.600
time, is a child, Genie.

00:51:30.600 --> 00:51:31.660
I'll show you.

00:51:31.660 --> 00:51:33.090
You'll see her in a movie
clip in a moment.

00:51:33.090 --> 00:51:37.800
They discovered her locked in
a room, in a house, from age

00:51:37.800 --> 00:51:39.510
20 months, approximately,
until they

00:51:39.510 --> 00:51:42.506
discovered her at 13.

00:51:42.506 --> 00:51:44.640
She almost spoke to no one.

00:51:44.640 --> 00:51:47.160
So the people taking care of
her locked her in a closet,

00:51:47.160 --> 00:51:49.832
basically, and did not talk
with her, or she had no

00:51:49.832 --> 00:51:51.550
language interactions
until she was

00:51:51.550 --> 00:51:52.945
discovered and rescued.

00:51:52.945 --> 00:51:55.865
And then they performed research
for years saying

00:51:55.865 --> 00:51:58.280
could she learn things, what
could she not learn.

00:51:58.280 --> 00:52:01.030
She got single words, for
example, but she never got

00:52:01.030 --> 00:52:02.190
particularly good at grammar.

00:52:02.190 --> 00:52:02.740
[INAUDIBLE]

00:52:02.740 --> 00:52:03.490
syntax.

00:52:03.490 --> 00:52:07.630
So a brutal way to test the idea
that what's acquirable

00:52:07.630 --> 00:52:08.780
later in life.

00:52:08.780 --> 00:52:11.330
And what, if you don't learn
it early in life, is

00:52:11.330 --> 00:52:13.170
impossible to acquire fully?

00:52:13.170 --> 00:52:13.980
So I'll end with this.

00:52:13.980 --> 00:52:15.790
We'll do this for five minutes
or seven minutes.

00:52:18.810 --> 00:52:20.070
I'll stop there.

00:52:20.070 --> 00:52:22.980
It's summarized in your book.

00:52:22.980 --> 00:52:25.720
Everybody knows about it because
it's this study you

00:52:25.720 --> 00:52:28.535
couldn't do ethically
in a million years.

00:52:28.535 --> 00:52:30.840
You would never do, which is if
you deprive a child of the

00:52:30.840 --> 00:52:32.780
critical period of language
exposure.

00:52:32.780 --> 00:52:35.220
And she never got lots of
aspects of language going.

00:52:35.220 --> 00:52:38.070
She got OK on a minimal
vocabulary,

00:52:38.070 --> 00:52:38.910
grew in other ways.

00:52:38.910 --> 00:52:41.235
But many of the things that,
from a language perspective,

00:52:41.235 --> 00:52:44.390
she never got because she missed
that critical period.

00:52:44.390 --> 00:52:47.000
So that's a bit of a sad story,
obviously, but have a

00:52:47.000 --> 00:52:48.815
good spring break and a
refreshing spring break.

00:52:48.815 --> 00:52:50.065
And I'll see you in 10 days.