1
00:00:00,090 --> 00:00:02,490
The following content is
provided under a Creative

2
00:00:02,490 --> 00:00:04,030
Commons license.

3
00:00:04,030 --> 00:00:06,330
Your support will help
MIT OpenCourseWare

4
00:00:06,330 --> 00:00:10,720
continue to offer high-quality
educational resources for free.

5
00:00:10,720 --> 00:00:13,320
To make a donation or
view additional materials

6
00:00:13,320 --> 00:00:17,280
from hundreds of MIT courses,
visit MIT OpenCourseWare

7
00:00:17,280 --> 00:00:20,430
at ocw.mit.edu.

8
00:00:20,430 --> 00:00:22,860
ABBY NOYCE: So we've
been talking this week

9
00:00:22,860 --> 00:00:24,790
about perception
of language, how

10
00:00:24,790 --> 00:00:27,090
you take all of this
information that's coming in,

11
00:00:27,090 --> 00:00:29,340
sort it out, figure out
the structure to it.

12
00:00:29,340 --> 00:00:33,564
We're going to talk today about
a slightly smaller subject,

13
00:00:33,564 --> 00:00:34,980
or something that
has been studied

14
00:00:34,980 --> 00:00:37,100
in less depth than
comprehension has,

15
00:00:37,100 --> 00:00:39,019
which is how you
produce language.

16
00:00:39,019 --> 00:00:40,560
You can do the same
thing in reverse.

17
00:00:40,560 --> 00:00:44,250
You can go from this kind of
non-semantic idea of something

18
00:00:44,250 --> 00:00:47,655
that you want to convey
to your listeners.

19
00:00:47,655 --> 00:00:49,590
So that's this kind
of message level,

20
00:00:49,590 --> 00:00:52,500
where you've got an
idea and from there,

21
00:00:52,500 --> 00:00:55,670
you select the words you're
going to use to convey it

22
00:00:55,670 --> 00:01:00,460
and you put together a
sentence structure for it.

23
00:01:00,460 --> 00:01:03,870
And from there, once you have
words in a sentence structure,

24
00:01:03,870 --> 00:01:08,040
then you need to pull up
the phonological information

25
00:01:08,040 --> 00:01:10,350
about the sentence, about
all of the words that

26
00:01:10,350 --> 00:01:13,260
are going into it.

27
00:01:13,260 --> 00:01:16,110
And finally, you pass all of
this information onto the motor

28
00:01:16,110 --> 00:01:18,180
control centers so
that you can actually

29
00:01:18,180 --> 00:01:19,305
articulate your sentence.

30
00:01:22,464 --> 00:01:24,770
So this is a fairly standard--

31
00:01:24,770 --> 00:01:27,890
This model shows that there
are four different levels

32
00:01:27,890 --> 00:01:31,700
of processing, and
these don't necessarily

33
00:01:31,700 --> 00:01:33,590
happen sequentially,
like you do all

34
00:01:33,590 --> 00:01:35,420
of your grammatical
encoding before any

35
00:01:35,420 --> 00:01:36,680
of your phonological encoding.

36
00:01:36,680 --> 00:01:39,380
There's some reasonably
good evidence

37
00:01:39,380 --> 00:01:41,090
that they interact a bit.

38
00:01:41,090 --> 00:01:45,110
And likewise, with phonological
encoding and articulation.

39
00:01:45,110 --> 00:01:50,330
But first, just to get
our brains working,

40
00:01:50,330 --> 00:01:53,780
we're going to have a moment
of design an experiment.

41
00:01:53,780 --> 00:01:58,700
So you're a cognitive
neuroscientist, or

42
00:01:58,700 --> 00:02:02,120
psycholinguist,
perhaps, and you believe

43
00:02:02,120 --> 00:02:05,330
that there are distinct
levels involved

44
00:02:05,330 --> 00:02:06,930
in the production of language.

45
00:02:06,930 --> 00:02:08,810
So when somebody
produces a sentence,

46
00:02:08,810 --> 00:02:11,150
they have to go through
different kinds of processing

47
00:02:11,150 --> 00:02:13,620
to do that.

48
00:02:13,620 --> 00:02:18,170
How might you prove
or disprove-- how

49
00:02:18,170 --> 00:02:21,110
might you test this hypothesis
that there are distinct

50
00:02:21,110 --> 00:02:24,470
phases in language production?

51
00:02:30,200 --> 00:02:31,060
Think about it.

52
00:02:31,060 --> 00:02:34,030
You can talk to your classmates.

53
00:02:34,030 --> 00:02:36,170
Things in your brain that
are involved in language.

54
00:02:36,170 --> 00:02:38,740
So we talked earlier
this week about,

55
00:02:38,740 --> 00:02:40,900
remember Wernicke had had
this idea that there were

56
00:02:40,900 --> 00:02:45,880
auditory images in Wernicke's
area and motor images

57
00:02:45,880 --> 00:02:46,840
and Broca's area.

58
00:02:46,840 --> 00:02:49,190
And so Wernicke's aphasia
versus Broca's aphasia,

59
00:02:49,190 --> 00:02:53,410
the difference was which of
these centers was harmed.

60
00:02:53,410 --> 00:02:57,220
And there's a lot of
flaws in that model.

61
00:02:57,220 --> 00:03:00,547
And this guy named
Marsel Mesulam modified

62
00:03:00,547 --> 00:03:02,630
it a little bit and said
that all language, again,

63
00:03:02,630 --> 00:03:05,080
depends on circuits,
on activities

64
00:03:05,080 --> 00:03:07,280
between different
language areas.

65
00:03:07,280 --> 00:03:09,430
And Mesulam was one
of the first ideas

66
00:03:09,430 --> 00:03:12,160
to point out that different
kinds of language production

67
00:03:12,160 --> 00:03:15,072
require different
kinds of processing.

68
00:03:15,072 --> 00:03:16,780
And then the example
he used is that if I

69
00:03:16,780 --> 00:03:18,850
ask you to do
something like name

70
00:03:18,850 --> 00:03:20,330
all the months in the year.

71
00:03:20,330 --> 00:03:20,830
Ready?

72
00:03:20,830 --> 00:03:21,150
Go.

73
00:03:21,150 --> 00:03:21,790
Come on.

74
00:03:21,790 --> 00:03:27,110
January, February, March,
April, May, June, July, August,

75
00:03:27,110 --> 00:03:30,274
September, October,
November, December.

76
00:03:30,274 --> 00:03:31,690
So this is something
that requires

77
00:03:31,690 --> 00:03:34,000
a lot less planning
and a lot less

78
00:03:34,000 --> 00:03:36,460
structuring than
actually producing

79
00:03:36,460 --> 00:03:39,890
a well-formed sentence does.

80
00:03:39,890 --> 00:03:41,740
And Mesulam
hypothesized that just

81
00:03:41,740 --> 00:03:44,470
rote motor connection like
this, or recite the alphabet,

82
00:03:44,470 --> 00:03:47,170
or recite your times tables,
or any of these things

83
00:03:47,170 --> 00:03:50,590
that you learned by heart
as a little kid, really

84
00:03:50,590 --> 00:03:52,930
only requires premotor
and motor areas up

85
00:03:52,930 --> 00:03:54,500
here in the frontal lobe.

86
00:03:54,500 --> 00:03:56,080
So we have our
primary motor cortex

87
00:03:56,080 --> 00:03:59,620
is this strip right along here,
and premotor is the blue stuff

88
00:03:59,620 --> 00:04:01,257
next to it.

89
00:04:01,257 --> 00:04:03,340
He pointed out, though,
when you're hearing words,

90
00:04:03,340 --> 00:04:06,790
this activates primarily
auditory cortex down here

91
00:04:06,790 --> 00:04:10,930
in the frontal lobe in
auditory association areas.

92
00:04:10,930 --> 00:04:13,900
Unimodal, so auditory
alone versus these kind

93
00:04:13,900 --> 00:04:17,170
of crossmodal representations,
when you're actually thinking

94
00:04:17,170 --> 00:04:18,490
about what you're hearing.

95
00:04:18,490 --> 00:04:22,000
If I say the word, I
don't know, baseball,

96
00:04:22,000 --> 00:04:24,670
you probably come up with
not just an auditory response

97
00:04:24,670 --> 00:04:26,680
to that, but all of
these other associations.

98
00:04:26,680 --> 00:04:28,721
You might have visual
memories that you associate

99
00:04:28,721 --> 00:04:32,506
with that or auditory
memories that

100
00:04:32,506 --> 00:04:33,880
aren't just the
sound of the word

101
00:04:33,880 --> 00:04:35,270
that you associate with that.

102
00:04:35,270 --> 00:04:37,260
So that would be like a
crossmodal association.

103
00:04:37,260 --> 00:04:39,970
But Mesulam says that when
you're just listening to words

104
00:04:39,970 --> 00:04:43,930
but not acting on them, then
most of what's activated

105
00:04:43,930 --> 00:04:46,165
is just auditory cortex.

106
00:04:49,310 --> 00:04:52,540
And Mesulam says when
you're producing words,

107
00:04:52,540 --> 00:04:55,480
then you end up seeing
activity all over the brain.

108
00:04:55,480 --> 00:04:56,380
You get motor stuff.

109
00:04:56,380 --> 00:04:58,354
You get word knowledge stuff.

110
00:04:58,354 --> 00:05:00,145
All of these different
pieces are involved.

111
00:05:05,020 --> 00:05:07,610
So here's an example of--

112
00:05:07,610 --> 00:05:11,110
this is some old-ish,
at this point, data--

113
00:05:11,110 --> 00:05:12,710
a PET scan.

114
00:05:12,710 --> 00:05:15,920
So these are participants
doing four different tasks.

115
00:05:15,920 --> 00:05:18,730
So they're looking
at a computer screen.

116
00:05:18,730 --> 00:05:23,792
And in the first
case, they would just

117
00:05:23,792 --> 00:05:25,750
show a word on the screen,
and the participants

118
00:05:25,750 --> 00:05:29,650
didn't have to do anything
but look at the word.

119
00:05:29,650 --> 00:05:31,630
And as you can
see, most of what's

120
00:05:31,630 --> 00:05:34,150
most active in that
condition is occipital.

121
00:05:34,150 --> 00:05:36,720
It's all that visual cortex.

122
00:05:36,720 --> 00:05:39,052
You can see there's a little
bit in the frontal lobes,

123
00:05:39,052 --> 00:05:40,510
a little bit in
the temporal lobes,

124
00:05:40,510 --> 00:05:41,926
but the areas that
are most active

125
00:05:41,926 --> 00:05:46,060
are the visual cortex in
the back of the brain there.

126
00:05:46,060 --> 00:05:48,530
The second case
was pretty similar,

127
00:05:48,530 --> 00:05:51,507
except instead of looking
at words, this was auditory.

128
00:05:51,507 --> 00:05:53,590
So you'd have headphones
on and look at the screen

129
00:05:53,590 --> 00:05:58,120
and you would hear words being
played into the headphones.

130
00:05:58,120 --> 00:05:59,770
And in this case,
it's mostly, as you

131
00:05:59,770 --> 00:06:04,220
might expect, temporal lobe
stuff that's most active.

132
00:06:04,220 --> 00:06:07,070
It's auditory cortex on
the sides of your head

133
00:06:07,070 --> 00:06:10,386
there in the temporal lobe.

134
00:06:10,386 --> 00:06:12,010
And one of the things
that's moderately

135
00:06:12,010 --> 00:06:14,230
interesting about that-- and I
don't have a good story for why

136
00:06:14,230 --> 00:06:16,646
this is-- is that a lot of the
frontal lobe activation you

137
00:06:16,646 --> 00:06:20,320
see in the looking-at-words case
goes away in the auditory case.

138
00:06:23,242 --> 00:06:25,572
AUDIENCE: [INAUDIBLE]?

139
00:06:25,572 --> 00:06:27,780
ABBY NOYCE: They were just
looking at a blank screen.

140
00:06:27,780 --> 00:06:29,760
I think it had like a
fixation point on it.

141
00:06:29,760 --> 00:06:33,730
"Keep looking at the
little dot on the screen."

142
00:06:33,730 --> 00:06:36,137
In the third case, they showed
them a word on the screen

143
00:06:36,137 --> 00:06:37,970
and participants had
to read the word aloud.

144
00:06:37,970 --> 00:06:39,950
They had to pronounce it.

145
00:06:39,950 --> 00:06:42,980
And as you can see, at this
point, a lot of the activity

146
00:06:42,980 --> 00:06:44,865
is suddenly in the motor cortex.

147
00:06:44,865 --> 00:06:46,220
I'm sure you all are shocked.

148
00:06:46,220 --> 00:06:51,260
So this is one piece of evidence
that articulation requires

149
00:06:51,260 --> 00:06:56,390
a different kind of activity
than simply understanding

150
00:06:56,390 --> 00:06:58,100
or perceiving
words, that you do,

151
00:06:58,100 --> 00:06:59,510
in fact, need motor activation.

152
00:06:59,510 --> 00:07:02,270
So this is that strip
between the frontal lobe

153
00:07:02,270 --> 00:07:05,250
and the parietal lobe,
where motor cortex is.

154
00:07:05,250 --> 00:07:08,250
And finally, they asked them
to do a word generation task.

155
00:07:08,250 --> 00:07:10,580
So they would show them
one word on a screen

156
00:07:10,580 --> 00:07:14,300
and ask the participant to
say a word that was related.

157
00:07:14,300 --> 00:07:17,450
So if the word that was
shown to them was "bike,"

158
00:07:17,450 --> 00:07:19,580
they might say "ride" or
they might say "helmet"

159
00:07:19,580 --> 00:07:21,750
or they might say "Cambridge."

160
00:07:21,750 --> 00:07:23,330
If the word on the
screen was "boat,"

161
00:07:23,330 --> 00:07:25,970
they might say "sail" and so on.

162
00:07:25,970 --> 00:07:28,264
So trying to find a
word that's associated

163
00:07:28,264 --> 00:07:29,430
with the word on the screen.

164
00:07:29,430 --> 00:07:32,810
So this suddenly requires
participants to actually use

165
00:07:32,810 --> 00:07:36,020
the semantics of the word in
a way that they didn't have to

166
00:07:36,020 --> 00:07:38,471
on any of the previous levels.

167
00:07:38,471 --> 00:07:40,970
If you're just reading it or
just hearing it or just reading

168
00:07:40,970 --> 00:07:44,127
it aloud, you don't really have
to think about what it means.

169
00:07:44,127 --> 00:07:46,460
Whereas if you're trying to
come up with a related word,

170
00:07:46,460 --> 00:07:48,740
you've got to tap into all
of your semantic knowledge

171
00:07:48,740 --> 00:07:51,710
about what that word represents.

172
00:07:51,710 --> 00:07:55,670
And what's interesting
about this is that you get,

173
00:07:55,670 --> 00:07:58,100
again, some motor
activation, and you

174
00:07:58,100 --> 00:08:00,440
get some auditory
activation, but you

175
00:08:00,440 --> 00:08:03,110
get a whole bunch of stuff going
on kind of in that premotor

176
00:08:03,110 --> 00:08:05,660
area, that area that kind
of coordinates and plans--

177
00:08:08,900 --> 00:08:11,480
utterances and other motor
things, but in this case,

178
00:08:11,480 --> 00:08:12,904
utterances.

179
00:08:12,904 --> 00:08:16,730
So this kind of imaging
scan of participants

180
00:08:16,730 --> 00:08:18,194
on tasks that
require them to think

181
00:08:18,194 --> 00:08:19,610
about language on
different levels

182
00:08:19,610 --> 00:08:23,390
is one piece of evidence
showing that there probably are

183
00:08:23,390 --> 00:08:25,066
different phases of processing.

184
00:08:30,559 --> 00:08:32,059
Another piece of
evidence that shows

185
00:08:32,059 --> 00:08:34,570
that there are different
stages in processing

186
00:08:34,570 --> 00:08:37,870
are the kinds of mistakes that
people make when they talk.

187
00:08:37,870 --> 00:08:40,419
We all got used to making
fun of President Bush

188
00:08:40,419 --> 00:08:42,516
because his command
of English language

189
00:08:42,516 --> 00:08:44,890
can be a little weak at times,
but the fact of the matter

190
00:08:44,890 --> 00:08:47,570
is that, if you
followed anybody,

191
00:08:47,570 --> 00:08:50,740
no matter how articulate they
are, around with microphones

192
00:08:50,740 --> 00:08:52,330
and wrote down
everything they said,

193
00:08:52,330 --> 00:08:53,560
you'd find tons of errors.

194
00:08:53,560 --> 00:08:56,380
We all make speech
errors anytime

195
00:08:56,380 --> 00:08:57,880
you're talking and
not just reading

196
00:08:57,880 --> 00:09:01,300
aloud, and sometimes even then.

197
00:09:01,300 --> 00:09:04,240
So one of the most common
types of speech errors

198
00:09:04,240 --> 00:09:06,370
are what are called
exchange errors,

199
00:09:06,370 --> 00:09:10,610
and you'll see these when people
are talking fast all the time.

200
00:09:10,610 --> 00:09:12,340
And you'll get both
word exchange errors

201
00:09:12,340 --> 00:09:15,280
like, "I wrote a
mother to my letter,"

202
00:09:15,280 --> 00:09:19,340
and you'll also see
sound exchange errors,

203
00:09:19,340 --> 00:09:22,910
phoneme exchange errors, and
these are called spoonerisms

204
00:09:22,910 --> 00:09:25,840
in honor of Archibald
Spooner, who

205
00:09:25,840 --> 00:09:29,740
was a professor at one of
the colleges in New York

206
00:09:29,740 --> 00:09:32,110
in like the early
20th century and was

207
00:09:32,110 --> 00:09:33,980
notorious for doing this.

208
00:09:33,980 --> 00:09:39,040
And there are many statements,
such as the one there,

209
00:09:39,040 --> 00:09:42,060
that have been attributed
to Monsieur Spooner,

210
00:09:42,060 --> 00:09:45,710
and he has given his name to
this kind of sound exchange

211
00:09:45,710 --> 00:09:46,720
error.

212
00:09:46,720 --> 00:09:49,390
"You've hissed all
my mystery lectures."

213
00:09:49,390 --> 00:09:51,107
"You have tasted
the whole worm."

214
00:09:54,516 --> 00:09:56,813
AUDIENCE: What's it
supposed to really be?

215
00:09:56,813 --> 00:09:57,438
AUDIENCE: Yeah.

216
00:09:57,438 --> 00:09:58,412
"Wasted the whole term."

217
00:09:58,412 --> 00:09:58,953
AUDIENCE: Oh.

218
00:09:58,953 --> 00:10:01,850
ABBY NOYCE: There you go.

219
00:10:01,850 --> 00:10:04,750
So sound exchange errors
are pretty common.

220
00:10:04,750 --> 00:10:06,940
I mean, I remember my
eighth grade earth science

221
00:10:06,940 --> 00:10:09,140
teacher sent the
entire class of us

222
00:10:09,140 --> 00:10:12,005
into about five minutes of
uncontrollable giggling.

223
00:10:12,005 --> 00:10:13,630
We were talking about
volcanoes, and he

224
00:10:13,630 --> 00:10:17,260
talked about how volcanoes would
release clouds of gash and ass.

225
00:10:17,260 --> 00:10:18,980
And of course, this
is eighth grade,

226
00:10:18,980 --> 00:10:21,250
and [GASPS] he said "ass."

227
00:10:21,250 --> 00:10:22,970
And there was much giggling.

228
00:10:22,970 --> 00:10:24,970
And it's just a straight
up sound exchange error

229
00:10:24,970 --> 00:10:26,350
like people make all the time.

230
00:10:29,180 --> 00:10:34,720
So let's look at some of these
different levels of processing

231
00:10:34,720 --> 00:10:39,070
that have to happen
in order to produce

232
00:10:39,070 --> 00:10:43,010
an utterance of some sort.

233
00:10:43,010 --> 00:10:46,090
So the message level is
probably pretty straightforward.

234
00:10:46,090 --> 00:10:49,300
You have some piece of
information, some idea,

235
00:10:49,300 --> 00:10:51,460
that you want to convey
to the people around you.

236
00:10:51,460 --> 00:10:56,150
How you move on from there
is a little bit harder.

237
00:10:56,150 --> 00:11:00,290
So one of the
things that happens

238
00:11:00,290 --> 00:11:02,110
is that there's this
word selection stage.

239
00:11:02,110 --> 00:11:05,480
So you're going from
this raw meaning idea

240
00:11:05,480 --> 00:11:07,430
and trying to find
the words that

241
00:11:07,430 --> 00:11:09,260
are necessary to represent it.

242
00:11:11,826 --> 00:11:13,700
Most likely, if we look
at the sort of models

243
00:11:13,700 --> 00:11:16,220
that we've been looking at
throughout this course, where

244
00:11:16,220 --> 00:11:18,847
you've got competing
activations,

245
00:11:18,847 --> 00:11:20,930
you can think of this as
like the meaning that you

246
00:11:20,930 --> 00:11:26,660
want to get across then is
going to activate different word

247
00:11:26,660 --> 00:11:29,420
representations depending on how
strongly related to the meaning

248
00:11:29,420 --> 00:11:30,710
they are.

249
00:11:30,710 --> 00:11:33,800
So if we think of the cortical
interconnections that we've

250
00:11:33,800 --> 00:11:36,740
considered all along.

251
00:11:36,740 --> 00:11:39,410
And sometimes, every
once in a while, you'll

252
00:11:39,410 --> 00:11:43,130
get errors that are a blend
of two different words.

253
00:11:43,130 --> 00:11:45,230
So sometimes when you
have two words that

254
00:11:45,230 --> 00:11:47,907
are both strongly activated
for a particular position

255
00:11:47,907 --> 00:11:50,240
in your sentence, you'll try
to say both of them at once

256
00:11:50,240 --> 00:11:51,200
and get all tangled up.

257
00:11:54,440 --> 00:11:56,870
The example in the
reading is, if you dropped

258
00:11:56,870 --> 00:11:59,275
a pen under somebody
in front of you's chair

259
00:11:59,275 --> 00:12:01,400
and you're trying to ask
them to pick it up for you

260
00:12:01,400 --> 00:12:03,937
and you've got both the word
"chair" and the word "seat"--

261
00:12:03,937 --> 00:12:05,520
you know, "My pen
is under your chair.

262
00:12:05,520 --> 00:12:06,603
My pen is under your seat.

263
00:12:06,603 --> 00:12:07,696
Can you get it for me?"

264
00:12:07,696 --> 00:12:09,320
But the "chair" and
the "seat" are both

265
00:12:09,320 --> 00:12:13,040
competing for that noun
slot in the middle there.

266
00:12:13,040 --> 00:12:17,690
And if they're both
equally strongly activated,

267
00:12:17,690 --> 00:12:20,840
they can both get passed down
to the phonological stage

268
00:12:20,840 --> 00:12:23,210
and you can end up
trying to pronounce both.

269
00:12:23,210 --> 00:12:25,640
And you'll say things like
"My pen is under your cheat--

270
00:12:25,640 --> 00:12:27,450
under your sair-- under--

271
00:12:27,450 --> 00:12:28,260
under your chair.

272
00:12:28,260 --> 00:12:29,314
Can I have it, please?"

273
00:12:29,314 --> 00:12:30,980
And I don't know if
anyone here has ever

274
00:12:30,980 --> 00:12:35,300
experienced that kind of
tangled up language production,

275
00:12:35,300 --> 00:12:39,980
but it's fairly common if you
start listening for this stuff.

276
00:12:44,450 --> 00:12:46,882
The other step in
grammatical encoding

277
00:12:46,882 --> 00:12:48,590
that's happening kind
of at the same time

278
00:12:48,590 --> 00:12:50,480
as you're trying to
find the words you need

279
00:12:50,480 --> 00:12:55,550
is you're trying
to understand how

280
00:12:55,550 --> 00:12:59,310
the syntax is put together--
what guides sentence syntax.

281
00:12:59,310 --> 00:13:01,070
And this is, at
least immediately,

282
00:13:01,070 --> 00:13:04,550
hard to think about because
all that you get to see of it

283
00:13:04,550 --> 00:13:05,330
is the output.

284
00:13:05,330 --> 00:13:07,160
It's what sentences
finally come up.

285
00:13:07,160 --> 00:13:11,600
You don't get to see any of
the intermediate processes.

286
00:13:11,600 --> 00:13:16,271
And so somebody did a
kind of clever study.

287
00:13:16,271 --> 00:13:20,180
They told participants they
were doing a memory experiment.

288
00:13:20,180 --> 00:13:23,930
And so they showed them
a string of simple images

289
00:13:23,930 --> 00:13:28,770
and said, hey, in order for
you to remember these better,

290
00:13:28,770 --> 00:13:31,010
we want you to say one
sentence about each image

291
00:13:31,010 --> 00:13:34,280
as it's presented to you.

292
00:13:34,280 --> 00:13:36,470
These were simple scenes.

293
00:13:36,470 --> 00:13:38,832
So you might have--

294
00:13:38,832 --> 00:13:40,040
I don't know if I showed you.

295
00:13:48,070 --> 00:13:49,400
Here's a house and a tree.

296
00:13:52,922 --> 00:13:54,880
And so for this sentence,
you could say either,

297
00:13:54,880 --> 00:13:57,050
the house is next to the
tree, or you could say,

298
00:13:57,050 --> 00:13:58,300
the tree is next to the house.

299
00:14:01,030 --> 00:14:06,190
And what these guys did
is, before each picture was

300
00:14:06,190 --> 00:14:08,080
presented, they
just put up a word

301
00:14:08,080 --> 00:14:10,300
and told subjects just
to read it out loud.

302
00:14:10,300 --> 00:14:16,270
And so for a scene like this,
they might put up a word like,

303
00:14:16,270 --> 00:14:21,010
oh, I don't know, "building"
versus a word like "pine."

304
00:14:21,010 --> 00:14:24,250
And what they were
hypothesizing was

305
00:14:24,250 --> 00:14:26,050
that, if your priming
word was a word that

306
00:14:26,050 --> 00:14:29,320
was related to the
house part, then

307
00:14:29,320 --> 00:14:33,590
when you tried to develop a
sentence about this picture,

308
00:14:33,590 --> 00:14:35,110
your ideas about
houses are going

309
00:14:35,110 --> 00:14:37,630
to be faster to
activate because they've

310
00:14:37,630 --> 00:14:40,349
been primed by the word
than your ideas about trees.

311
00:14:40,349 --> 00:14:41,890
And so you would
expect to see people

312
00:14:41,890 --> 00:14:44,767
say sentences that were like,
the house is next to the tree.

313
00:14:44,767 --> 00:14:46,600
They'd pull up the house
part first and then

314
00:14:46,600 --> 00:14:49,180
fill the rest of it in later.

315
00:14:49,180 --> 00:14:51,310
Whereas if the priming
sentence was something

316
00:14:51,310 --> 00:14:54,530
like pine or maple, then
your tree representations

317
00:14:54,530 --> 00:14:58,570
would be more active and easier
and faster to be pulled up.

318
00:14:58,570 --> 00:15:00,640
And so you would start
developing a sentence

319
00:15:00,640 --> 00:15:03,400
structure that let
the tree part go first

320
00:15:03,400 --> 00:15:07,810
and fill in later
with the other half

321
00:15:07,810 --> 00:15:13,817
of the image with the house, so
that you would see people say,

322
00:15:13,817 --> 00:15:15,400
the tree is next to
the house, if they

323
00:15:15,400 --> 00:15:19,050
were primed with one of
those tree-related words.

324
00:15:19,050 --> 00:15:22,490
And pretty much
what they found is

325
00:15:22,490 --> 00:15:24,670
that the part of the scene
that was primed generally

326
00:15:24,670 --> 00:15:28,540
came first in the
sentences people spoke.

327
00:15:28,540 --> 00:15:31,180
And the part of the scene that
was not related to the priming

328
00:15:31,180 --> 00:15:32,160
word came second.

329
00:15:32,160 --> 00:15:33,940
So you'd see people
putting things

330
00:15:33,940 --> 00:15:37,600
into active versus passive
voice depending on which

331
00:15:37,600 --> 00:15:39,029
part of the scene was first--

332
00:15:39,029 --> 00:15:41,320
not so much for this image,
but an image that shows one

333
00:15:41,320 --> 00:15:43,090
object acting on another object.

334
00:15:45,700 --> 00:15:51,310
So this idea that word
selection and sentence structure

335
00:15:51,310 --> 00:15:54,100
happen kind of at the same time
and how fast you can pull up

336
00:15:54,100 --> 00:15:56,162
the words you want is
going to affect how

337
00:15:56,162 --> 00:15:57,370
you structure your sentences.

338
00:15:57,370 --> 00:15:59,770
If you can get one
word right away,

339
00:15:59,770 --> 00:16:01,990
then you're going to build
the rest of the sentence

340
00:16:01,990 --> 00:16:04,677
to accommodate putting
that word at the beginning

341
00:16:04,677 --> 00:16:06,760
and filling in later with
the other words that are

342
00:16:06,760 --> 00:16:08,110
a little bit slower to come up.

343
00:16:23,530 --> 00:16:25,590
So back to these
ideas about stages

344
00:16:25,590 --> 00:16:28,800
that are involved in
language production.

345
00:16:28,800 --> 00:16:32,070
We talked about grammatical
encoding earlier.

346
00:16:32,070 --> 00:16:33,750
Let's talk about
this next stage down,

347
00:16:33,750 --> 00:16:37,680
this phonological encoding.

348
00:16:37,680 --> 00:16:39,330
When people are
speaking fluently,

349
00:16:39,330 --> 00:16:43,206
they usually produce about
three words every second.

350
00:16:43,206 --> 00:16:45,080
And so for each of these
words, the speaker's

351
00:16:45,080 --> 00:16:47,370
got to be able to pull up
the phonological information

352
00:16:47,370 --> 00:16:50,220
about it-- the information
about what pattern of phonemes,

353
00:16:50,220 --> 00:16:55,950
what pattern of sounds, makes up
that word, how you stress them,

354
00:16:55,950 --> 00:16:58,950
how it maybe changes when
it's next to other words

355
00:16:58,950 --> 00:17:02,341
in a particular order,
all of that information.

356
00:17:06,190 --> 00:17:11,949
And one thing that you tend to
see happening when people fail

357
00:17:11,949 --> 00:17:13,740
at pulling up this
phonological information

358
00:17:13,740 --> 00:17:16,690
is what's call a "tip
of the tongue" state.

359
00:17:16,690 --> 00:17:22,050
When somebody maybe reads you
a definition and asks you,

360
00:17:22,050 --> 00:17:23,280
what's that word?

361
00:17:23,280 --> 00:17:24,690
And you know it.

362
00:17:24,690 --> 00:17:27,960
You know it's a word you
know, and the word just

363
00:17:27,960 --> 00:17:29,580
won't come to you, or the name.

364
00:17:29,580 --> 00:17:30,330
Who's that person?

365
00:17:30,330 --> 00:17:30,996
Oh, I know them.

366
00:17:30,996 --> 00:17:34,290
They are-- and you
just can't get it.

367
00:17:34,290 --> 00:17:36,221
Anyone ever have
this happen to them?

368
00:17:36,221 --> 00:17:38,380
AUDIENCE: All the time.

369
00:17:38,380 --> 00:17:39,130
ABBY NOYCE: Right.

370
00:17:39,130 --> 00:17:40,630
So it's a "tip of
the tongue" state.

371
00:17:40,630 --> 00:17:45,430
And in this model, this
"tip of the tongue" scenario

372
00:17:45,430 --> 00:17:48,550
is explained by when you know
all of the semantic stuff

373
00:17:48,550 --> 00:17:50,314
about the word.

374
00:17:50,314 --> 00:17:51,730
If it's a name
that you can't get,

375
00:17:51,730 --> 00:17:53,140
you know which person
you're talking about.

376
00:17:53,140 --> 00:17:55,348
You can tell somebody all
sorts of things about them,

377
00:17:55,348 --> 00:17:58,060
like if they were in your
history class last semester

378
00:17:58,060 --> 00:18:02,290
or whatever, but you can't
come up with the name.

379
00:18:02,290 --> 00:18:04,150
And this is generally
believed that when

380
00:18:04,150 --> 00:18:07,240
the connection between the
semantic representation

381
00:18:07,240 --> 00:18:10,900
and the phonological information
is blocked for whatever reason,

382
00:18:10,900 --> 00:18:14,110
that you can't go
from the semantic step

383
00:18:14,110 --> 00:18:16,750
to the phonological step.

384
00:18:16,750 --> 00:18:19,480
"Tip of the tongue" states
are more common for moderately

385
00:18:19,480 --> 00:18:21,280
uncommon words.

386
00:18:21,280 --> 00:18:24,210
So for really common,
everyday, short words,

387
00:18:24,210 --> 00:18:26,470
"tip of the tongue"
states are pretty rare.

388
00:18:26,470 --> 00:18:28,510
For longer, more
precise words, words

389
00:18:28,510 --> 00:18:31,060
that you don't see so
often, they're more common.

390
00:18:31,060 --> 00:18:33,070
Usually when you have
a "what's that word?"

391
00:18:33,070 --> 00:18:35,530
it's for a word that's
not one of the most

392
00:18:35,530 --> 00:18:38,720
common in the language.

393
00:18:38,720 --> 00:18:41,620
And there have been a
few cases of patients,

394
00:18:41,620 --> 00:18:44,830
usually following some kind
of brain injury, brain trauma,

395
00:18:44,830 --> 00:18:47,950
where they have this "tip
of the tongue" phenomenon

396
00:18:47,950 --> 00:18:48,590
all the time.

397
00:18:48,590 --> 00:18:50,240
They have a great
deal of difficulty.

398
00:18:50,240 --> 00:18:53,757
For every word they try and
come up with, they just can't--

399
00:18:53,757 --> 00:18:55,090
they know what they want to say.

400
00:18:55,090 --> 00:18:57,610
They've got the semantic
representation right there,

401
00:18:57,610 --> 00:19:00,771
and they just can't find the
phonological representation

402
00:19:00,771 --> 00:19:01,270
for it.

403
00:19:04,633 --> 00:19:07,450
Yeah, tip of the tongue.

404
00:19:07,450 --> 00:19:09,920
So this is another
kind of example

405
00:19:09,920 --> 00:19:13,540
of a stage where the
sentence construction

406
00:19:13,540 --> 00:19:18,820
stage, the grammatical encoding
stage, of producing language

407
00:19:18,820 --> 00:19:23,890
is working OK, and then this
later phonological encoding is

408
00:19:23,890 --> 00:19:25,090
not working.

409
00:19:25,090 --> 00:19:27,670
So another piece of evidence
that these might, in fact,

410
00:19:27,670 --> 00:19:32,200
be separate phases in language
processing, language creation,

411
00:19:32,200 --> 00:19:35,930
language production, something.

412
00:19:35,930 --> 00:19:40,074
So do these things ever
interact with each other?

413
00:19:40,074 --> 00:19:41,740
The easiest way of
looking at this model

414
00:19:41,740 --> 00:19:45,190
is saying that sentences start
out as abstract meanings, which

415
00:19:45,190 --> 00:19:47,380
then are passed to this
grammatical encoding

416
00:19:47,380 --> 00:19:51,190
level, which finds words and
builds a structure, which then

417
00:19:51,190 --> 00:19:53,140
passes them to the
phonological level, which

418
00:19:53,140 --> 00:19:56,260
finds sounds, which then passes
them to this articulation

419
00:19:56,260 --> 00:19:58,390
machine.

420
00:19:58,390 --> 00:20:00,100
But it turns out
there's actually

421
00:20:00,100 --> 00:20:02,500
at least some feedback
going on between the levels,

422
00:20:02,500 --> 00:20:04,546
or it seems to be.

423
00:20:04,546 --> 00:20:07,900
And the best evidence
for this at the moment

424
00:20:07,900 --> 00:20:10,480
is that we talked about
word exchange errors.

425
00:20:10,480 --> 00:20:13,870
"I wrote a mother to my letter."

426
00:20:13,870 --> 00:20:17,620
And these occur-- if
there was no interaction

427
00:20:17,620 --> 00:20:21,370
between the grammatical levels
and the phonological levels,

428
00:20:21,370 --> 00:20:27,730
then you'd expect to see that
word exchange errors happen

429
00:20:27,730 --> 00:20:30,430
equally often between
pairs of words,

430
00:20:30,430 --> 00:20:35,860
disregarding whether they have
any phonological similarity.

431
00:20:35,860 --> 00:20:38,620
If you actually document what
kinds of word exchange errors

432
00:20:38,620 --> 00:20:40,540
people tend to make,
this isn't true.

433
00:20:40,540 --> 00:20:44,080
The word exchange errors
people make occur more often

434
00:20:44,080 --> 00:20:46,540
in words that have some kind
of similar phoneme pattern.

435
00:20:46,540 --> 00:20:50,470
So "mother" and "letter" both
have a stressed first syllable,

436
00:20:50,470 --> 00:20:53,310
and then that last syllable
with just the E-R on it

437
00:20:53,310 --> 00:20:55,500
that's unstressed.

438
00:20:55,500 --> 00:20:57,619
And patterns like that
are what you tend to see.

439
00:20:57,619 --> 00:20:59,410
So there's some amount
of feedback going on

440
00:20:59,410 --> 00:21:00,284
between these levels.

441
00:21:04,080 --> 00:21:07,050
Articulation, of
course, this last stage.

442
00:21:07,050 --> 00:21:09,340
Articulation is mostly just
a matter of motor control.

443
00:21:09,340 --> 00:21:13,030
It's saying, OK, I've got a
phonological representation.

444
00:21:13,030 --> 00:21:17,550
How do I turn that into
actual, specific movements

445
00:21:17,550 --> 00:21:21,360
of my mouth, my tongue, my
lips, all of this stuff that's

446
00:21:21,360 --> 00:21:23,190
actually articulating it?

447
00:21:23,190 --> 00:21:24,960
So here's our good
old brain diagram.

448
00:21:24,960 --> 00:21:28,680
Again, primary
motor cortex here.

449
00:21:28,680 --> 00:21:30,270
And it's been really
well documented

450
00:21:30,270 --> 00:21:32,490
at this point that
is kind of laid out

451
00:21:32,490 --> 00:21:35,100
over a strip along motor
cortex are regions that

452
00:21:35,100 --> 00:21:36,360
respond to different things.

453
00:21:36,360 --> 00:21:41,670
So toes, ankles, hip,
knee, trunk, arm.

454
00:21:41,670 --> 00:21:43,880
Notice there's a lot of
cortex devoted to your hands.

455
00:21:43,880 --> 00:21:46,560
That's probably
not too shocking.

456
00:21:46,560 --> 00:21:49,280
Fine control over hands is
really important for humans.

457
00:21:49,280 --> 00:21:52,510
It's important for a lot
of things that we do.

458
00:21:52,510 --> 00:21:55,140
There's also a lot of it
connected to, like, lips

459
00:21:55,140 --> 00:21:56,910
and jaw and tongue.

460
00:21:56,910 --> 00:22:01,410
Again, this ability to have
really fine and precise control

461
00:22:01,410 --> 00:22:04,437
over these things is important.

462
00:22:04,437 --> 00:22:07,020
I'm always kind of startled at
just how much motor cortex goes

463
00:22:07,020 --> 00:22:09,870
to the swallowing and throat
control, which makes sense,

464
00:22:09,870 --> 00:22:11,370
but it's not something
that we think

465
00:22:11,370 --> 00:22:12,744
of as requiring
a lot of control.

466
00:22:20,480 --> 00:22:23,780
So motor cortex
happening right up there.

467
00:22:23,780 --> 00:22:27,110
Remember, if we go back to--

468
00:22:27,110 --> 00:22:29,769
what was his name?

469
00:22:29,769 --> 00:22:32,060
The guy we were talking about
at the beginning of class

470
00:22:32,060 --> 00:22:33,800
who had a model of
how language depends

471
00:22:33,800 --> 00:22:39,999
on different parts
of the motor cortex.

472
00:22:39,999 --> 00:22:41,540
And I'm totally
blanking on his name.

473
00:22:41,540 --> 00:22:43,780
"Tip of the tongue" moment.

474
00:22:43,780 --> 00:22:45,675
Starts with an M. What was it?

475
00:22:45,675 --> 00:22:48,360
AUDIENCE: Mesulam.

476
00:22:48,360 --> 00:22:49,730
ABBY NOYCE: Mesulam, OK.

477
00:22:49,730 --> 00:22:50,900
Yeah, him.

478
00:22:50,900 --> 00:22:53,300
And he was saying that,
depending on what you're doing,

479
00:22:53,300 --> 00:22:57,710
some kinds of articulation
basically just require

480
00:22:57,710 --> 00:22:58,320
motor control.

481
00:22:58,320 --> 00:23:00,590
Others have to be kind of
coordinated from further up.

482
00:23:03,530 --> 00:23:06,110
So we've got our motor cortex.

483
00:23:06,110 --> 00:23:08,150
So the primary motor
cortex, the strip

484
00:23:08,150 --> 00:23:11,210
that's right along
the sulcus there,

485
00:23:11,210 --> 00:23:13,790
right up against
the parietal lobe,

486
00:23:13,790 --> 00:23:17,360
controls mostly fine
motor coordination,

487
00:23:17,360 --> 00:23:18,820
very fine control of movement.

488
00:23:18,820 --> 00:23:22,400
There's actually nerve fibers
that go from primary cortex

489
00:23:22,400 --> 00:23:23,880
all the way out to the muscles.

490
00:23:23,880 --> 00:23:26,660
This is like direct muscular
control for fine motion.

491
00:23:29,630 --> 00:23:32,690
And then next to it is
this kind of premotor area.

492
00:23:32,690 --> 00:23:35,240
And the premotor area seems to
be most involved in setting up

493
00:23:35,240 --> 00:23:38,480
sequences of actions,
especially in response

494
00:23:38,480 --> 00:23:40,200
to perceptual information.

495
00:23:40,200 --> 00:23:44,450
So for example, if I
was to toss something

496
00:23:44,450 --> 00:23:46,790
across the room to a student.

497
00:23:46,790 --> 00:23:48,490
Sarah, catch.

498
00:23:48,490 --> 00:23:50,810
And Sarah is-- very nice.

499
00:23:50,810 --> 00:23:54,110
So Sarah is taking in all of
this perceptual information,

500
00:23:54,110 --> 00:23:56,420
where she sees the pen
coming towards her,

501
00:23:56,420 --> 00:24:00,230
and coordinating this plan that
involves moving her hands up

502
00:24:00,230 --> 00:24:03,080
at the right time and
in the right place

503
00:24:03,080 --> 00:24:05,750
in order to be able to catch
the pen when it gets to her.

504
00:24:05,750 --> 00:24:07,880
And that's coordinated in part--

505
00:24:07,880 --> 00:24:09,980
setting up that
sequence is coordinated

506
00:24:09,980 --> 00:24:13,049
through the premotor
area and the cerebellum.

507
00:24:13,049 --> 00:24:14,840
The cerebellum has a
lot to do with getting

508
00:24:14,840 --> 00:24:18,950
this kind of smooth coordination
so that our muscles are all

509
00:24:18,950 --> 00:24:21,650
working at the right time
relative to one another

510
00:24:21,650 --> 00:24:25,190
so that we aren't
moving all jerkily.

511
00:24:25,190 --> 00:24:27,740
And then there's this kind
of supplementary motor

512
00:24:27,740 --> 00:24:34,070
area that is involved in more
self-directed action planning.

513
00:24:34,070 --> 00:24:36,902
So premotor is really
working with responding

514
00:24:36,902 --> 00:24:39,110
to your immediate environment,
taking that perception

515
00:24:39,110 --> 00:24:42,170
and building a
sequence from there.

516
00:24:42,170 --> 00:24:44,660
Supplementary motor
area is involved

517
00:24:44,660 --> 00:24:46,070
in longer-term planning.

518
00:24:51,050 --> 00:24:53,319
If I was like, I
have a motor goal

519
00:24:53,319 --> 00:24:54,860
of getting to that
window, that would

520
00:24:54,860 --> 00:24:57,230
be coordinated by the
supplementary motor area.

521
00:24:57,230 --> 00:25:01,040
And then that
premotor area would

522
00:25:01,040 --> 00:25:03,920
be more involved in helping
me walk around obstacles

523
00:25:03,920 --> 00:25:06,110
and make sure I don't
walk into anything

524
00:25:06,110 --> 00:25:08,570
on my way of getting there.

525
00:25:08,570 --> 00:25:12,820
So all of these are going to
be involved in articulation.

526
00:25:12,820 --> 00:25:18,170
Someone did a study in
monkeys, training them to--

527
00:25:18,170 --> 00:25:21,950
no, what am I thinking of?

528
00:25:21,950 --> 00:25:23,420
Yeah, this was a
study in monkeys,

529
00:25:23,420 --> 00:25:25,750
but it wasn't a directly
articulation-related one.

530
00:25:25,750 --> 00:25:27,083
Someone did a study in monkeys--

531
00:25:29,680 --> 00:25:31,810
because monkeys
don't have language--

532
00:25:31,810 --> 00:25:34,550
where they had
either trained them

533
00:25:34,550 --> 00:25:39,680
to push a set of three
buttons in a certain pattern

534
00:25:39,680 --> 00:25:42,740
or train them to look at how
the buttons lit up and then

535
00:25:42,740 --> 00:25:44,960
repeat the pattern by
pushing the buttons.

536
00:25:44,960 --> 00:25:48,860
So one of these was a
task where the monkey had

537
00:25:48,860 --> 00:25:50,254
to know the plan in advance.

538
00:25:50,254 --> 00:25:51,920
One of these was a
task where the monkey

539
00:25:51,920 --> 00:25:55,280
had to respond to
environmental input

540
00:25:55,280 --> 00:25:57,230
to perceptual information.

541
00:25:57,230 --> 00:26:00,460
And this is one of the really
solid pieces of evidence

542
00:26:00,460 --> 00:26:02,840
for the difference between
supplementary motor

543
00:26:02,840 --> 00:26:05,060
and premotor in terms
of what they do.

544
00:26:05,060 --> 00:26:07,639
When the monkeys were
working entirely from memory,

545
00:26:07,639 --> 00:26:09,680
then it was mostly their
supplementary motor area

546
00:26:09,680 --> 00:26:10,910
that was active.

547
00:26:10,910 --> 00:26:12,770
When they were
working in response

548
00:26:12,770 --> 00:26:15,311
to what was shown to them right
then, it was mostly premotor.

549
00:26:28,030 --> 00:26:30,580
One other thing about
articulation before we move on

550
00:26:30,580 --> 00:26:35,710
is just that, because
our face is right up

551
00:26:35,710 --> 00:26:38,440
here close to the brain,
a lot of the nerves that

552
00:26:38,440 --> 00:26:41,279
control the muscles that
are involved in articulation

553
00:26:41,279 --> 00:26:43,570
don't run through the spinal
cord the way a lot of them

554
00:26:43,570 --> 00:26:46,510
do for, say,
controlling your hands.

555
00:26:46,510 --> 00:26:48,340
A lot of it is these
direct cranial nerves

556
00:26:48,340 --> 00:26:50,500
that come right from the brain.

557
00:26:50,500 --> 00:26:52,990
They don't go out and through
the spinal cord and back.

558
00:26:52,990 --> 00:26:55,724
So the way that the circuit
works for controlling

559
00:26:55,724 --> 00:26:57,640
facial stuff is different
than for controlling

560
00:26:57,640 --> 00:27:00,610
a lot of the rest of your body.

561
00:27:00,610 --> 00:27:01,950
Moving on very quickly.

562
00:27:01,950 --> 00:27:04,360
The other context in which
we tend to produce language

563
00:27:04,360 --> 00:27:05,980
is writing things.

564
00:27:08,710 --> 00:27:11,950
How we produce spoken language
is not well understood.

565
00:27:11,950 --> 00:27:17,620
How we produce written language
is even less well understood.

566
00:27:17,620 --> 00:27:19,634
But there are some
key differences

567
00:27:19,634 --> 00:27:21,050
between producing
written language

568
00:27:21,050 --> 00:27:23,830
and between speaking
language that I just wanted

569
00:27:23,830 --> 00:27:25,570
to highlight for you guys.

570
00:27:28,157 --> 00:27:30,490
Unlike when you talk-- you're
usually talking to people,

571
00:27:30,490 --> 00:27:34,000
with people, with people
immediately around you--

572
00:27:34,000 --> 00:27:36,550
when you're writing, it's
usually just you, usually just

573
00:27:36,550 --> 00:27:41,020
your stream of language
that's being put down.

574
00:27:41,020 --> 00:27:43,510
You aren't building it into
like a conversation or anything

575
00:27:43,510 --> 00:27:45,280
like that.

576
00:27:45,280 --> 00:27:46,960
If you analyze the
syntax that's used

577
00:27:46,960 --> 00:27:49,210
in people's writings
versus a syntax that's

578
00:27:49,210 --> 00:27:52,120
used when they are
speaking, it's more complex.

579
00:27:52,120 --> 00:27:54,520
You'll see more recursion,
more subordinate clauses, more

580
00:27:54,520 --> 00:27:56,927
nested phrases, all of that.

581
00:27:56,927 --> 00:27:59,260
And this seems to be easier
to follow when reading, too.

582
00:27:59,260 --> 00:28:02,450
So this make sense
on both sides.

583
00:28:02,450 --> 00:28:04,450
And when you're writing,
you get to futz with it

584
00:28:04,450 --> 00:28:06,490
and change it after you can
write it down and be like, no,

585
00:28:06,490 --> 00:28:07,490
that doesn't make sense.

586
00:28:07,490 --> 00:28:08,530
Let me go back.

587
00:28:08,530 --> 00:28:09,910
You don't get to do that
when you're talking.

588
00:28:09,910 --> 00:28:11,785
If you say something
that doesn't make sense,

589
00:28:11,785 --> 00:28:15,090
you're kind of stuck
with it being out there.