1
00:00:10,200 --> 00:00:12,750
NANCY KANWISHER: So we won't
get through the whole attention

2
00:00:12,750 --> 00:00:13,110
lecture.

3
00:00:13,110 --> 00:00:13,860
I saw this coming.

4
00:00:13,860 --> 00:00:16,027
I just felt like that last
bit was important enough.

5
00:00:16,027 --> 00:00:17,430
We'll cut as needed.

6
00:00:17,430 --> 00:00:19,380
But I do want to
talk about attention.

7
00:00:19,380 --> 00:00:22,710
And let me start by
sharing with you some

8
00:00:22,710 --> 00:00:26,430
of the key ideas about
what attention is.

9
00:00:26,430 --> 00:00:29,580
OK, so to get into the mode
of thinking about this,

10
00:00:29,580 --> 00:00:32,220
let's consider the
following question.

11
00:00:32,220 --> 00:00:34,530
How do you feel about
people driving while talking

12
00:00:34,530 --> 00:00:36,700
on their cell phones?

13
00:00:36,700 --> 00:00:39,540
Smart, not smart?

14
00:00:39,540 --> 00:00:42,990
And while you're thinking about
that, also consider the case--

15
00:00:42,990 --> 00:00:45,480
like let's suppose
that you have had

16
00:00:45,480 --> 00:00:47,760
a hands-free setup,
so you don't have

17
00:00:47,760 --> 00:00:50,190
to be looking down
on your cell phone

18
00:00:50,190 --> 00:00:55,740
or typing away or
holding your cell phone.

19
00:00:55,740 --> 00:00:57,360
Maybe that's OK.

20
00:00:57,360 --> 00:00:58,710
What do you think?

21
00:00:58,710 --> 00:01:01,740
Is there no problem with driving
while talking on your cell

22
00:01:01,740 --> 00:01:03,570
phone, provided you're
not looking down

23
00:01:03,570 --> 00:01:04,920
and pecking away at it?

24
00:01:09,440 --> 00:01:10,040
Yes?

25
00:01:10,040 --> 00:01:10,790
What do you think?

26
00:01:13,078 --> 00:01:14,620
I want to know what
you really think.

27
00:01:14,620 --> 00:01:18,190
I know what all the PR says.

28
00:01:18,190 --> 00:01:19,840
Is it fine?

29
00:01:19,840 --> 00:01:20,980
Yes?

30
00:01:20,980 --> 00:01:23,950
AUDIENCE: I would say it depends
on whether you're driving--

31
00:01:23,950 --> 00:01:26,500
Like if the road is clear and
it's like your usual path,

32
00:01:26,500 --> 00:01:28,600
like driving is pretty
automatic at that point.

33
00:01:28,600 --> 00:01:30,820
But it it's like high
traffic or a new area,

34
00:01:30,820 --> 00:01:33,050
then I think it would have
a significant affects,

35
00:01:33,050 --> 00:01:34,510
so you shouldn't.

36
00:01:34,510 --> 00:01:35,650
NANCY KANWISHER: Totally.

37
00:01:35,650 --> 00:01:38,620
My rule is, if I am
talking on my cell phone,

38
00:01:38,620 --> 00:01:41,380
certainly, if I'm turning
left, I put the damn cell phone

39
00:01:41,380 --> 00:01:42,160
on my lap.

40
00:01:42,160 --> 00:01:45,430
And I tell whoever it
is, I can't talk to you

41
00:01:45,430 --> 00:01:47,050
and turn left at the same time.

42
00:01:47,050 --> 00:01:49,240
All right, turning left is
the hardest thing we do.

43
00:01:49,240 --> 00:01:51,113
Self-driving cars
can't turn left.

44
00:01:51,113 --> 00:01:52,030
Have I mentioned this?

45
00:01:52,030 --> 00:01:53,800
How does a self-driving
car turn left?

46
00:01:53,800 --> 00:01:55,780
It turns right three times.

47
00:01:55,780 --> 00:01:56,350
Why is that?

48
00:01:56,350 --> 00:01:58,840
Because turning left
is social cognition.

49
00:01:58,840 --> 00:02:01,060
It's like, do they see me?

50
00:02:01,060 --> 00:02:02,145
Do they know I'm here?

51
00:02:02,145 --> 00:02:03,520
Are they looking
at my indicator?

52
00:02:03,520 --> 00:02:05,145
Do they realize I'm
going to turn left?

53
00:02:05,145 --> 00:02:05,710
This is hard.

54
00:02:05,710 --> 00:02:06,790
This is social cognition.

55
00:02:06,790 --> 00:02:08,380
Nobody's mastered that yet.

56
00:02:08,380 --> 00:02:11,110
Anyway, it's hard for us too.

57
00:02:11,110 --> 00:02:14,140
Anyway, the real
question here is not

58
00:02:14,140 --> 00:02:17,410
for me to lecture you about
driving but to think about,

59
00:02:17,410 --> 00:02:20,650
why is that a problem?

60
00:02:20,650 --> 00:02:24,700
Why can't you talk to a person
and pay attention to driving?

61
00:02:24,700 --> 00:02:27,970
What is the big deal, right?

62
00:02:27,970 --> 00:02:32,260
And so the intuition is we have
some kind of limited processing

63
00:02:32,260 --> 00:02:33,190
ability.

64
00:02:33,190 --> 00:02:36,430
OK, we're not like super
supercomputers who can just

65
00:02:36,430 --> 00:02:38,230
do millions of things at once.

66
00:02:38,230 --> 00:02:41,150
We have some kind of limit in
how much stuff we can handle,

67
00:02:41,150 --> 00:02:41,650
OK?

68
00:02:41,650 --> 00:02:46,210
And I think everybody
will have this intuition.

69
00:02:46,210 --> 00:02:48,730
You can't think about lots
of different things at once.

70
00:02:48,730 --> 00:02:52,780
I like to think of this as the
toaster model of cognition.

71
00:02:52,780 --> 00:02:55,570
That is, you plug in the
toaster, and the lights dim,

72
00:02:55,570 --> 00:02:56,260
right?

73
00:02:56,260 --> 00:02:58,030
So if everything's
on the same circuit,

74
00:02:58,030 --> 00:03:01,810
there's some kind of unitary
thing, resource that's limited.

75
00:03:01,810 --> 00:03:05,020
And if some more of it goes
here, less if it goes here.

76
00:03:05,020 --> 00:03:07,272
OK, so obviously we don't
have simple circuits

77
00:03:07,272 --> 00:03:08,230
like that in our brain.

78
00:03:08,230 --> 00:03:11,530
And the relevant
scarce commodity

79
00:03:11,530 --> 00:03:13,150
isn't just electricity.

80
00:03:13,150 --> 00:03:14,080
It's not electrons.

81
00:03:14,080 --> 00:03:18,010
It's some other kind of thing
we don't totally understand.

82
00:03:18,010 --> 00:03:22,960
But there's still some sense in
which of our mental processes

83
00:03:22,960 --> 00:03:24,580
are on the same circuit.

84
00:03:24,580 --> 00:03:25,810
OK?

85
00:03:25,810 --> 00:03:29,120
I'm assuming everybody is
sharing this intuition here.

86
00:03:29,120 --> 00:03:30,278
So think about this.

87
00:03:30,278 --> 00:03:32,320
How many people feel like
you can listen to music

88
00:03:32,320 --> 00:03:34,458
and read difficult
stuff at the same time?

89
00:03:34,458 --> 00:03:35,500
Raise your-- I'm curious.

90
00:03:35,500 --> 00:03:38,230
How many people feel
like they can do that?

91
00:03:38,230 --> 00:03:40,090
Only a few.

92
00:03:40,090 --> 00:03:41,980
That's so interesting.

93
00:03:41,980 --> 00:03:42,730
Sort of.

94
00:03:42,730 --> 00:03:45,310
Yeah, I can't do it
at all, like at all,

95
00:03:45,310 --> 00:03:47,740
like even background
music that I don't even

96
00:03:47,740 --> 00:03:49,900
care that much about.

97
00:03:49,900 --> 00:03:51,730
Someday-- I mean, I
think these are really

98
00:03:51,730 --> 00:03:53,090
stunning individual differences.

99
00:03:53,090 --> 00:03:54,830
I don't know if there's
any good work on it.

100
00:03:54,830 --> 00:03:55,580
I haven't seen it.

101
00:03:55,580 --> 00:03:58,190
But I think it's
really interesting.

102
00:03:58,190 --> 00:03:58,900
Some people can.

103
00:03:58,900 --> 00:03:59,650
Some people can't.

104
00:03:59,650 --> 00:04:02,740
I don't know what's
up with that.

105
00:04:02,740 --> 00:04:07,343
How about recognizing faces
and scenes at the same time?

106
00:04:07,343 --> 00:04:08,010
Can you do that?

107
00:04:08,010 --> 00:04:08,460
Yeah?

108
00:04:08,460 --> 00:04:10,543
How many people feel like
you can recognize a face

109
00:04:10,543 --> 00:04:12,095
and a scene at the same time?

110
00:04:12,095 --> 00:04:15,210
All right.

111
00:04:15,210 --> 00:04:17,850
In fact, Michael Cohen,
who gave that lecture

112
00:04:17,850 --> 00:04:21,990
on brain-computer interfaces
a month ago or so,

113
00:04:21,990 --> 00:04:25,230
showed some beautiful
work that, actually, you

114
00:04:25,230 --> 00:04:29,700
are better at recognizing a
face and a scene presented

115
00:04:29,700 --> 00:04:35,550
simultaneously than two
faces or two scenes.

116
00:04:35,550 --> 00:04:38,430
Can you think why that might be?

117
00:04:38,430 --> 00:04:39,720
Yeah, Isabelle?

118
00:04:39,720 --> 00:04:40,977
AUDIENCE: Context.

119
00:04:40,977 --> 00:04:42,060
NANCY KANWISHER: Say more.

120
00:04:44,880 --> 00:04:50,960
AUDIENCE: If you see somebody,
even a familiar face,

121
00:04:50,960 --> 00:04:52,780
you can either--

122
00:04:52,780 --> 00:04:55,970
it helps identify them as,
oh, I know that person.

123
00:04:55,970 --> 00:04:57,412
Then, obviously, they're there.

124
00:04:57,412 --> 00:04:58,370
NANCY KANWISHER: Right.

125
00:04:58,370 --> 00:05:00,710
So they can go together
into a thing, maybe chunk

126
00:05:00,710 --> 00:05:01,985
it as a thing or something.

127
00:05:01,985 --> 00:05:02,860
That's a good answer.

128
00:05:02,860 --> 00:05:05,210
It wasn't the one
I was fishing for.

129
00:05:05,210 --> 00:05:07,250
Yeah, Jimmy?

130
00:05:07,250 --> 00:05:12,950
AUDIENCE: If people and
places are separate things

131
00:05:12,950 --> 00:05:15,980
that your brain calculates,
if it's two faces,

132
00:05:15,980 --> 00:05:19,250
you have to calculate this
thing with a face number one

133
00:05:19,250 --> 00:05:21,230
before going to face number two.

134
00:05:21,230 --> 00:05:24,140
Whereas if it's [INAUDIBLE]

135
00:05:24,140 --> 00:05:25,640
NANCY KANWISHER: Exactly.

136
00:05:25,640 --> 00:05:26,360
Exactly.

137
00:05:26,360 --> 00:05:27,470
And you do, right?

138
00:05:27,470 --> 00:05:28,820
You have an FFA and a PPA.

139
00:05:28,820 --> 00:05:30,620
And they're friends, right?

140
00:05:30,620 --> 00:05:33,470
So what Michael showed
is not just that you

141
00:05:33,470 --> 00:05:36,260
can recognize a face and
hold it in working memory,

142
00:05:36,260 --> 00:05:39,710
a face and a scene better
than two faces or two scenes,

143
00:05:39,710 --> 00:05:42,320
but the degree to which
you have that cost

144
00:05:42,320 --> 00:05:44,000
of doing two things
in one category

145
00:05:44,000 --> 00:05:46,220
rather than spreading
over two categories

146
00:05:46,220 --> 00:05:50,210
is linearly proportional to
how different the activation

147
00:05:50,210 --> 00:05:51,710
patterns are for
those categories

148
00:05:51,710 --> 00:05:53,390
in the ventral visual pathway.

149
00:05:53,390 --> 00:05:56,040
So faces and scenes are totally
different from each other.

150
00:05:56,040 --> 00:06:00,860
And so you get a big benefit
for separating your two items

151
00:06:00,860 --> 00:06:03,530
across faces and scenes.

152
00:06:03,530 --> 00:06:05,660
Other things that are more
similar in their pattern

153
00:06:05,660 --> 00:06:08,960
of response in the ventral
pathway, like faces and bodies,

154
00:06:08,960 --> 00:06:10,620
you get a smaller benefit.

155
00:06:10,620 --> 00:06:11,120
OK?

156
00:06:11,120 --> 00:06:12,495
So it's consistent
with this idea

157
00:06:12,495 --> 00:06:15,230
that, to the degree that we
have separate processors,

158
00:06:15,230 --> 00:06:21,020
you can do, to some extent,
processing in parallel, OK?

159
00:06:21,020 --> 00:06:25,127
It doesn't explain why not
everybody can read and listen

160
00:06:25,127 --> 00:06:26,210
to music at the same time.

161
00:06:26,210 --> 00:06:27,752
Because, as I argued
a few weeks ago,

162
00:06:27,752 --> 00:06:29,543
these things don't
overlap at all.

163
00:06:29,543 --> 00:06:30,710
So there are many mysteries.

164
00:06:30,710 --> 00:06:33,660
But there's some
intuition there.

165
00:06:33,660 --> 00:06:35,910
OK, to get a little
more intuition

166
00:06:35,910 --> 00:06:38,533
about limited processing
capacity, can you guys see?

167
00:06:38,533 --> 00:06:40,200
When I was sitting
over there last time,

168
00:06:40,200 --> 00:06:42,100
I couldn't see
the screen at all.

169
00:06:42,100 --> 00:06:43,360
Can you guys see it?

170
00:06:43,360 --> 00:06:43,860
It's OK?

171
00:06:43,860 --> 00:06:44,400
All right.

172
00:06:44,400 --> 00:06:45,185
Yeah?

173
00:06:45,185 --> 00:06:51,063
AUDIENCE: Do you think
the problem is [INAUDIBLE]

174
00:06:51,063 --> 00:06:52,230
NANCY KANWISHER: Not for me.

175
00:06:52,230 --> 00:06:55,120
Instrumental music, no
lyrics still a problem.

176
00:06:55,120 --> 00:06:55,833
So I don't know.

177
00:06:55,833 --> 00:06:57,750
I mean, I'm sure there's
a literature on this.

178
00:06:57,750 --> 00:06:58,740
I just don't happen to know it.

179
00:06:58,740 --> 00:07:00,610
I'm just trying to
share the intuition.

180
00:07:00,610 --> 00:07:03,240
So I don't-- it's not
the only thing for me.

181
00:07:06,520 --> 00:07:09,220
OK, so what I'm going to do--
this is a super low-tech demo.

182
00:07:09,220 --> 00:07:11,613
I'm going to show you an
array of colored letters.

183
00:07:11,613 --> 00:07:13,030
Grab a piece of
paper or something

184
00:07:13,030 --> 00:07:14,800
where you can write
down a few letters.

185
00:07:14,800 --> 00:07:17,350
And I'm just going to flash
up one array very quickly.

186
00:07:17,350 --> 00:07:20,740
And I want you to write down
as many of the blue letters

187
00:07:20,740 --> 00:07:21,820
as you can.

188
00:07:21,820 --> 00:07:22,510
OK, ready?

189
00:07:30,275 --> 00:07:31,400
There's only a few of them.

190
00:07:31,400 --> 00:07:32,380
So you can.

191
00:07:32,380 --> 00:07:32,980
OK, ready?

192
00:07:32,980 --> 00:07:33,550
Here we go.

193
00:07:36,277 --> 00:07:37,110
OK, write them down.

194
00:07:39,820 --> 00:07:41,320
Last year, everyone
got all of them.

195
00:07:41,320 --> 00:07:43,560
So I tried to go a
little faster this time.

196
00:07:43,560 --> 00:07:45,938
OK.

197
00:07:45,938 --> 00:07:47,230
OK, we're going to do it again.

198
00:07:47,230 --> 00:07:48,000
Ready?

199
00:07:48,000 --> 00:07:49,110
There's going to
be another display,

200
00:07:49,110 --> 00:07:50,730
and you're going to write
down all the blue letters.

201
00:07:50,730 --> 00:07:51,720
Everyone ready?

202
00:07:51,720 --> 00:07:52,230
Here we go.

203
00:07:58,520 --> 00:08:00,830
OK, just wanted to
share the intuition.

204
00:08:00,830 --> 00:08:04,050
The probability-- did
anyone miss the N here?

205
00:08:04,050 --> 00:08:06,755
OK, how many people got the N?

206
00:08:06,755 --> 00:08:07,880
Nobody's going to admit it.

207
00:08:10,780 --> 00:08:13,150
I wanted a few people
to miss the N, right?

208
00:08:13,150 --> 00:08:14,210
I really did zip through.

209
00:08:14,210 --> 00:08:16,317
I probably cheated and
made this one go faster.

210
00:08:16,317 --> 00:08:16,900
You missed it?

211
00:08:16,900 --> 00:08:17,647
OK, thank you.

212
00:08:17,647 --> 00:08:18,730
I'm sure a few others did.

213
00:08:18,730 --> 00:08:24,760
OK, anyway, the point
is that, if you do this

214
00:08:24,760 --> 00:08:26,500
properly in a lab,
the probability

215
00:08:26,500 --> 00:08:30,340
of detecting it here is
much less than there, right?

216
00:08:30,340 --> 00:08:33,315
So what does this show?

217
00:08:33,315 --> 00:08:34,940
It shows that we have
limited capacity.

218
00:08:34,940 --> 00:08:37,429
We can't process all
of those blue letters

219
00:08:37,429 --> 00:08:39,770
in parallel, right?

220
00:08:39,770 --> 00:08:41,909
When there's a single
one, you get it just fine.

221
00:08:41,909 --> 00:08:44,120
But when there's a bunch,
you don't necessarily

222
00:08:44,120 --> 00:08:45,620
get it, right?

223
00:08:45,620 --> 00:08:47,090
OK.

224
00:08:47,090 --> 00:08:49,700
It also shows that we
have ability to select

225
00:08:49,700 --> 00:08:51,530
the information, right?

226
00:08:51,530 --> 00:08:53,695
So you weren't bothered
by the red letters.

227
00:08:53,695 --> 00:08:55,820
If I'd asked you right
after about the red letters,

228
00:08:55,820 --> 00:08:58,970
you probably couldn't report
any of them at all, right?

229
00:08:58,970 --> 00:09:00,800
It's a whole field
that studies that.

230
00:09:00,800 --> 00:09:02,450
We won't get into that.

231
00:09:02,450 --> 00:09:04,310
OK.

232
00:09:04,310 --> 00:09:06,740
So in fact, the
probability of reporting

233
00:09:06,740 --> 00:09:10,220
the N in these two displays
is independent of the number

234
00:09:10,220 --> 00:09:11,420
of red letters.

235
00:09:11,420 --> 00:09:12,930
They just don't matter.

236
00:09:12,930 --> 00:09:16,040
They are not entering into that
limitation, whatever it is.

237
00:09:16,040 --> 00:09:17,630
The limitation is
only for the ones

238
00:09:17,630 --> 00:09:23,400
you're paying attention
to, the blue ones, OK?

239
00:09:23,400 --> 00:09:29,370
OK, so why is our
mental capacity limited?

240
00:09:29,370 --> 00:09:30,630
OK?

241
00:09:30,630 --> 00:09:32,100
Nobody really knows an answer.

242
00:09:32,100 --> 00:09:34,890
And I actually think this is
a really unsatisfying part

243
00:09:34,890 --> 00:09:37,780
of attention research because
it sort of seems obvious.

244
00:09:37,780 --> 00:09:39,120
But I think it's not.

245
00:09:39,120 --> 00:09:42,120
So sometime in the
next 10 or 20 years,

246
00:09:42,120 --> 00:09:45,780
some smart computational person
will analyze this well and get

247
00:09:45,780 --> 00:09:46,960
a more satisfying answer.

248
00:09:46,960 --> 00:09:48,418
But right now,
here's where we are.

249
00:09:50,760 --> 00:09:55,290
The standard story is, well,
we have only so many neurons

250
00:09:55,290 --> 00:09:56,370
and so much capacity.

251
00:09:56,370 --> 00:09:59,730
And you can't do
everything at once, right?

252
00:09:59,730 --> 00:10:02,280
So we can't process everything
in the whole visual field

253
00:10:02,280 --> 00:10:04,790
at once.

254
00:10:04,790 --> 00:10:06,920
That's the standard story.

255
00:10:06,920 --> 00:10:09,070
But the reason I find
this unsatisfying is like,

256
00:10:09,070 --> 00:10:10,070
why the hell not, right?

257
00:10:10,070 --> 00:10:13,310
We have all this parallel stuff
or the whole visual field,

258
00:10:13,310 --> 00:10:15,590
at least in the first
few stages of processing.

259
00:10:15,590 --> 00:10:18,440
And we have quite a degree
of parallelism after that.

260
00:10:18,440 --> 00:10:20,270
So I think that's
the standard story.

261
00:10:20,270 --> 00:10:23,480
But I'm just marking that I
don't find it very satisfying.

262
00:10:23,480 --> 00:10:26,090
Yeah?

263
00:10:26,090 --> 00:10:28,490
A second answer, which
I think is also not

264
00:10:28,490 --> 00:10:31,970
totally satisfying but also
a little bit right at least,

265
00:10:31,970 --> 00:10:35,310
is that, typically, you're
only going to do one action.

266
00:10:35,310 --> 00:10:35,810
Right?

267
00:10:35,810 --> 00:10:37,185
You're going around
in the world.

268
00:10:37,185 --> 00:10:40,080
Let's forget colored letters
on a display in a lab.

269
00:10:40,080 --> 00:10:43,190
Let's think about a person
walking around in a busy city

270
00:10:43,190 --> 00:10:45,150
street doing something, right?

271
00:10:45,150 --> 00:10:48,200
There's typically only
one or a very small number

272
00:10:48,200 --> 00:10:50,450
of things you're going to
do next, like walk down

273
00:10:50,450 --> 00:10:52,220
the sidewalk and
not bump into people

274
00:10:52,220 --> 00:10:54,740
or pick up this object, right?

275
00:10:54,740 --> 00:10:58,430
You don't need to act towards
all the things in your world.

276
00:10:58,430 --> 00:11:01,160
And in fact, if you
had to, so given

277
00:11:01,160 --> 00:11:03,230
that you're only going
to do one thing, why

278
00:11:03,230 --> 00:11:05,720
distract the whole rest of
your motor-planning system

279
00:11:05,720 --> 00:11:07,970
by feeding it all this
information that's is just

280
00:11:07,970 --> 00:11:09,470
going to clutter
it with garbage?

281
00:11:09,470 --> 00:11:12,185
Why not give it just the
information it needs?

282
00:11:12,185 --> 00:11:13,560
OK, so that's the
standard story.

283
00:11:13,560 --> 00:11:18,410
It's very squishy and vague
but hopefully intuitive.

284
00:11:18,410 --> 00:11:22,340
And there's some nice,
compelling examples

285
00:11:22,340 --> 00:11:23,430
that illustrate this.

286
00:11:23,430 --> 00:11:25,730
So there's a fish
called the pike fish

287
00:11:25,730 --> 00:11:28,460
that preys on smaller
fish called sticklebacks.

288
00:11:28,460 --> 00:11:29,510
OK?

289
00:11:29,510 --> 00:11:34,640
And if you stick a pike fish
in a tank with sticklebacks,

290
00:11:34,640 --> 00:11:36,500
it will catch the
first stickleback

291
00:11:36,500 --> 00:11:39,620
faster if there's only
one in the tank with it

292
00:11:39,620 --> 00:11:42,690
than if there are 10
in the tank with it.

293
00:11:42,690 --> 00:11:43,190
OK?

294
00:11:43,190 --> 00:11:45,710
You might think, you've
got 10 to choose from.

295
00:11:45,710 --> 00:11:47,810
You'll get a fish faster.

296
00:11:47,810 --> 00:11:50,420
But 10 is more
distracting, right?

297
00:11:50,420 --> 00:11:52,700
And so maybe our
action-planning systems also

298
00:11:52,700 --> 00:11:55,430
prefer the single stickleback.

299
00:11:55,430 --> 00:11:57,440
Again, there's no
computational precision.

300
00:11:57,440 --> 00:12:00,110
I'm just sharing
intuitions with you.

301
00:12:00,110 --> 00:12:03,890
OK, so let me give you some
more evidence that there really

302
00:12:03,890 --> 00:12:09,377
are significant capacity
limits in perception

303
00:12:09,377 --> 00:12:11,960
and that, in fact, there's a lot
of stuff right in front of us

304
00:12:11,960 --> 00:12:15,800
that comes right onto our
retina that we don't see.

305
00:12:15,800 --> 00:12:17,000
OK?

306
00:12:17,000 --> 00:12:19,350
So I'm going to
show you an example.

307
00:12:19,350 --> 00:12:21,740
I'm going to show you a
picture for just a few seconds.

308
00:12:21,740 --> 00:12:22,970
And I want you to
just look at it.

309
00:12:22,970 --> 00:12:24,770
And then I'm going to ask
you some questions about it.

310
00:12:24,770 --> 00:12:25,070
OK?

311
00:12:25,070 --> 00:12:25,640
Here we go.

312
00:12:28,710 --> 00:12:33,300
OK, so how rich a
percept did you get?

313
00:12:33,300 --> 00:12:37,335
Did you see lots of stuff
and get lots of detail?

314
00:12:37,335 --> 00:12:38,710
Or do you just
feel like you just

315
00:12:38,710 --> 00:12:42,400
have the vaguest sense of a few
buildings, a street, that's it?

316
00:12:42,400 --> 00:12:47,450
How many feel like they got
a lot of pretty good detail?

317
00:12:47,450 --> 00:12:48,222
No one does.

318
00:12:48,222 --> 00:12:49,055
Or everyone's bored.

319
00:12:49,055 --> 00:12:51,578
I don't know which.

320
00:12:51,578 --> 00:12:54,200
AUDIENCE: What do
you mean good detail?

321
00:12:54,200 --> 00:12:56,450
NANCY KANWISHER: The colors
of buildings, the presence

322
00:12:56,450 --> 00:12:59,840
of objects, details on
the architectural styles,

323
00:12:59,840 --> 00:13:02,878
where an awning was,
that kind of thing.

324
00:13:02,878 --> 00:13:08,228
AUDIENCE: I feel like
I got [INAUDIBLE]

325
00:13:08,228 --> 00:13:10,520
NANCY KANWISHER: Well, most
people looking at this feel

326
00:13:10,520 --> 00:13:10,760
like--

327
00:13:10,760 --> 00:13:11,900
I mean, who knows what it means?

328
00:13:11,900 --> 00:13:13,692
I mean, we're just
sharing intuitions here.

329
00:13:13,692 --> 00:13:15,495
But most feel like,
that was pretty rich.

330
00:13:15,495 --> 00:13:16,370
I have a sense of it.

331
00:13:16,370 --> 00:13:18,162
I have a sense of maybe
what country that's

332
00:13:18,162 --> 00:13:19,610
in and what kind
of stuff is there

333
00:13:19,610 --> 00:13:21,482
and what kind the style is.

334
00:13:21,482 --> 00:13:23,690
You get a feel for the place,
all that kind of thing.

335
00:13:23,690 --> 00:13:26,930
Maybe not every damn
detail, but lots, right?

336
00:13:26,930 --> 00:13:28,880
So the general intuition
most people have

337
00:13:28,880 --> 00:13:31,400
is a fair amount of detail.

338
00:13:31,400 --> 00:13:33,050
OK?

339
00:13:33,050 --> 00:13:34,550
Well, let's look
at that some more.

340
00:13:34,550 --> 00:13:38,870
This is actually a very,
very heated topic right now.

341
00:13:38,870 --> 00:13:41,600
Me and Michael Cohen published
a paper a couple of years

342
00:13:41,600 --> 00:13:44,030
ago called The Bandwidth of
Perception, which is an effort

343
00:13:44,030 --> 00:13:46,940
to grapple with this question
of how much information is there

344
00:13:46,940 --> 00:13:48,710
in your current
percept right now.

345
00:13:48,710 --> 00:13:51,360
And there are many different
perspectives on this.

346
00:13:51,360 --> 00:13:54,810
Let's find out a little
more with a further demo.

347
00:13:54,810 --> 00:13:57,380
So what I'm going
to do is I'm going

348
00:13:57,380 --> 00:14:01,040
to show you that picture again.

349
00:14:01,040 --> 00:14:03,690
And it's going to flash
on a number of times.

350
00:14:03,690 --> 00:14:05,150
And each time it
flashes on, there

351
00:14:05,150 --> 00:14:07,050
might be something
that's different.

352
00:14:07,050 --> 00:14:10,640
So take notes on any
differences you might detect.

353
00:14:10,640 --> 00:14:12,685
OK?

354
00:14:12,685 --> 00:14:13,185
OK.

355
00:14:40,380 --> 00:14:42,280
OK, what things changed?

356
00:14:42,280 --> 00:14:42,780
Yeah?

357
00:14:42,780 --> 00:14:45,030
AUDIENCE: There was a woman
walking down the sidewalk.

358
00:14:45,030 --> 00:14:46,290
NANCY KANWISHER: And?

359
00:14:46,290 --> 00:14:47,598
AUDIENCE: That's all I got.

360
00:14:47,598 --> 00:14:49,140
NANCY KANWISHER:
OK, but she changed.

361
00:14:49,140 --> 00:14:51,840
What changed about her?

362
00:14:51,840 --> 00:14:54,793
AUDIENCE: Other
than her clothes?

363
00:14:54,793 --> 00:14:55,710
NANCY KANWISHER: Yeah.

364
00:14:55,710 --> 00:14:57,450
Over the successive
presentations,

365
00:14:57,450 --> 00:14:59,190
was she there sometimes
and not others?

366
00:14:59,190 --> 00:14:59,820
Did she change?

367
00:14:59,820 --> 00:15:01,500
Yeah, she appeared
or disappeared.

368
00:15:01,500 --> 00:15:02,010
OK, good.

369
00:15:02,010 --> 00:15:03,394
What else, Isabel?

370
00:15:03,394 --> 00:15:04,846
AUDIENCE: The awnings changed.

371
00:15:04,846 --> 00:15:06,835
The buildings changed colors.

372
00:15:06,835 --> 00:15:07,960
NANCY KANWISHER: Very good.

373
00:15:07,960 --> 00:15:08,710
Very good.

374
00:15:08,710 --> 00:15:12,790
How many people saw
an awning change?

375
00:15:12,790 --> 00:15:14,560
Maybe half of you.

376
00:15:14,560 --> 00:15:15,370
OK.

377
00:15:15,370 --> 00:15:17,878
What else changed?

378
00:15:17,878 --> 00:15:19,238
AUDIENCE: The car.

379
00:15:19,238 --> 00:15:20,530
AUDIENCE: The model of the car.

380
00:15:20,530 --> 00:15:21,640
NANCY KANWISHER: The
model of the car.

381
00:15:21,640 --> 00:15:22,390
Very good.

382
00:15:22,390 --> 00:15:22,900
Yeah?

383
00:15:22,900 --> 00:15:25,300
AUDIENCE: I feel like
the shops changed.

384
00:15:25,300 --> 00:15:27,930
But I don't know when or how.

385
00:15:27,930 --> 00:15:30,590
NANCY KANWISHER: OK.

386
00:15:30,590 --> 00:15:31,090
OK.

387
00:15:31,090 --> 00:15:32,770
You want to see
how they changed?

388
00:15:32,770 --> 00:15:34,600
OK, here's how it ended.

389
00:15:34,600 --> 00:15:36,460
Here's how it started.

390
00:15:36,460 --> 00:15:38,140
Are you surprised how much?

391
00:15:38,140 --> 00:15:40,600
Raise your hand if you're
surprised how much changed.

392
00:15:40,600 --> 00:15:43,410
Yeah, a lot, right?

393
00:15:43,410 --> 00:15:45,910
OK, so what does that tell us?

394
00:15:45,910 --> 00:15:49,680
It tells us that
either the sense

395
00:15:49,680 --> 00:15:52,950
we have that we really saw a
lot of what's going on there

396
00:15:52,950 --> 00:15:53,460
was wrong.

397
00:15:53,460 --> 00:15:54,960
We didn't see as
much as we thought.

398
00:15:54,960 --> 00:15:56,910
Because if we saw as
much as we thought,

399
00:15:56,910 --> 00:15:59,343
we should notice massive
changes like that.

400
00:15:59,343 --> 00:16:01,260
If you didn't see the
awning, look over there.

401
00:16:01,260 --> 00:16:04,560
I mean, how could-- look
at the color changes.

402
00:16:04,560 --> 00:16:06,630
Pretty major, right?

403
00:16:06,630 --> 00:16:09,360
Or we perceive all that
stuff in the instant,

404
00:16:09,360 --> 00:16:12,255
and it just goes poof by the
time the next one comes along.

405
00:16:12,255 --> 00:16:14,130
That's one of the things
the field is finding

406
00:16:14,130 --> 00:16:16,088
about those things are
very hard to tell apart,

407
00:16:16,088 --> 00:16:17,640
not impossible, but hard.

408
00:16:17,640 --> 00:16:19,710
Yeah.

409
00:16:19,710 --> 00:16:23,190
OK, so most people feel
like, wow, much more changed

410
00:16:23,190 --> 00:16:24,450
than I thought.

411
00:16:24,450 --> 00:16:28,230
I can't resist one
more hilarious demo.

412
00:16:28,230 --> 00:16:30,990
So have you already
seen this like in 9.00?

413
00:16:30,990 --> 00:16:33,080
How many people have
seen this before?

414
00:16:33,080 --> 00:16:34,503
Oh, I don't want to bore you.

415
00:16:34,503 --> 00:16:35,670
Do you mind seeing it again?

416
00:16:35,670 --> 00:16:36,690
We could skip it.

417
00:16:36,690 --> 00:16:38,110
It's kind of fun.

418
00:16:38,110 --> 00:16:38,610
Sorry?

419
00:16:38,610 --> 00:16:38,880
Do it?

420
00:16:38,880 --> 00:16:39,420
OK.

421
00:16:39,420 --> 00:16:41,440
OK.

422
00:16:41,440 --> 00:16:41,940
OK.

423
00:16:41,940 --> 00:16:44,310
So you're just going to
watch this video and just

424
00:16:44,310 --> 00:16:45,905
track things
because there may be

425
00:16:45,905 --> 00:16:47,280
changes happening
here and there,

426
00:16:47,280 --> 00:16:49,090
and you want to notice them, OK?

427
00:16:49,090 --> 00:16:49,590
Here we go.

428
00:16:49,590 --> 00:16:53,173
[MUSIC PLAYING]

429
00:16:53,173 --> 00:16:53,840
[VIDEO PLAYBACK]

430
00:16:53,840 --> 00:16:57,350
- Clearly, somebody in this
room murdered Lord Smythe,

431
00:16:57,350 --> 00:17:00,860
who, at precisely
3:34 this afternoon,

432
00:17:00,860 --> 00:17:04,770
was brutally bludgeoned to
death with a blunt instrument.

433
00:17:04,770 --> 00:17:07,970
I want each of you to
tell me your whereabouts

434
00:17:07,970 --> 00:17:13,290
at precisely the time that
this dastardly deed took place.

435
00:17:13,290 --> 00:17:17,069
- I was polishing the brass
in the master bedroom.

436
00:17:17,069 --> 00:17:21,380
- I was buttering his lordship's
scones below stairs, sir.

437
00:17:21,380 --> 00:17:24,800
- I was planting my petunias
in the potting shed.

438
00:17:24,800 --> 00:17:29,340
- Constable, arrest Lady Smythe.

439
00:17:29,340 --> 00:17:30,140
- How did you know?

440
00:17:30,140 --> 00:17:32,250
- Madam, as any
horticulturist will tell you,

441
00:17:32,250 --> 00:17:35,060
one does not plant
petunias until May is out.

442
00:17:35,060 --> 00:17:37,010
Take her away.

443
00:17:37,010 --> 00:17:39,080
It's just a matter
of observation.

444
00:17:39,080 --> 00:17:42,440
The real question is,
how observant were you?

445
00:17:51,380 --> 00:17:54,920
Clearly, somebody in this
room murdered Lord Smythe,

446
00:17:54,920 --> 00:17:58,490
who at precisely
3:34 this afternoon,

447
00:17:58,490 --> 00:18:02,280
was brutally bludgeoned to
death with a blunt instrument.

448
00:18:02,280 --> 00:18:05,540
I want each of you to
tell me your whereabouts

449
00:18:05,540 --> 00:18:08,587
at precisely the time that
this dastardly deed took place.

450
00:18:08,587 --> 00:18:09,170
[END PLAYBACK]

451
00:18:09,170 --> 00:18:10,070
NANCY KANWISHER:
Totally different guy.

452
00:18:10,070 --> 00:18:10,737
[VIDEO PLAYBACK]

453
00:18:10,737 --> 00:18:14,510
- I was polishing the brass
in the master bedroom.

454
00:18:14,510 --> 00:18:18,890
- I was buttering his lordship's
scones below stairs, sir.

455
00:18:18,890 --> 00:18:22,280
- I was planting my petunias
in the potting shed.

456
00:18:22,280 --> 00:18:25,520
- Constable, arrest Lady Smythe.

457
00:18:29,091 --> 00:18:29,674
[END PLAYBACK]

458
00:18:29,674 --> 00:18:31,216
NANCY KANWISHER:
It's actually an ad.

459
00:18:31,216 --> 00:18:33,110
But it doesn't hurt
to get its message.

460
00:18:36,480 --> 00:18:42,270
All right, it's a British ad
with an important message.

461
00:18:42,270 --> 00:18:44,815
But it's a pretty impressive
demo too, isn't it?

462
00:18:44,815 --> 00:18:46,440
How many people feel
like lots of stuff

463
00:18:46,440 --> 00:18:48,720
changed that they didn't notice?

464
00:18:48,720 --> 00:18:50,940
Yeah, that's intuition.

465
00:18:50,940 --> 00:18:54,630
So the idea here is that all
that stuff hits your retina.

466
00:18:54,630 --> 00:18:57,300
All of it kind of gets
processed to some degree.

467
00:18:57,300 --> 00:19:00,810
But it's amazing how much
of it goes unnoticed.

468
00:19:00,810 --> 00:19:02,060
OK?

469
00:19:02,060 --> 00:19:03,020
All right, oops.

470
00:19:03,020 --> 00:19:04,490
Sorry.

471
00:19:04,490 --> 00:19:08,050
OK, you might think, OK, is
this something that only happens

472
00:19:08,050 --> 00:19:09,300
when it doesn't really matter?

473
00:19:09,300 --> 00:19:10,790
OK, you were
looking for changes.

474
00:19:10,790 --> 00:19:12,740
But your life
didn't depend on it.

475
00:19:12,740 --> 00:19:14,900
What about commercial pilots?

476
00:19:14,900 --> 00:19:16,610
So here's a classic
study where they

477
00:19:16,610 --> 00:19:19,940
brought in actual
real commercial pilots

478
00:19:19,940 --> 00:19:23,210
with thousands of hours
of flying experience.

479
00:19:23,210 --> 00:19:25,910
And they had them fly
in a flight simulator.

480
00:19:25,910 --> 00:19:30,860
They had them land a plane on a
runway in the flight simulator

481
00:19:30,860 --> 00:19:33,080
under foggy conditions.

482
00:19:33,080 --> 00:19:34,850
And I forget what percent.

483
00:19:34,850 --> 00:19:37,045
I have to cheat and
look at my notes.

484
00:19:37,045 --> 00:19:38,270
It doesn't say.

485
00:19:38,270 --> 00:19:38,810
It does.

486
00:19:38,810 --> 00:19:39,768
I'm just not seeing it.

487
00:19:39,768 --> 00:19:43,220
Some large percent
of the pilots never

488
00:19:43,220 --> 00:19:46,880
saw this plane sitting
right there on the runway.

489
00:19:46,880 --> 00:19:49,880
They landed the planes
in the simulator right

490
00:19:49,880 --> 00:19:51,142
through that plane.

491
00:19:51,142 --> 00:19:52,850
And when they were
shown it subsequently,

492
00:19:52,850 --> 00:19:54,433
they were shocked
and couldn't believe

493
00:19:54,433 --> 00:19:55,820
they didn't see it, right?

494
00:19:55,820 --> 00:19:57,430
They were using a
heads-up display

495
00:19:57,430 --> 00:19:58,805
that tells them
their orientation

496
00:19:58,805 --> 00:19:59,900
to where the runway is.

497
00:19:59,900 --> 00:20:01,700
And they're paying
attention to this stuff,

498
00:20:01,700 --> 00:20:02,720
landing on the runway.

499
00:20:02,720 --> 00:20:06,830
And they just do not even see
that very relevant thing, OK?

500
00:20:06,830 --> 00:20:10,160
So it's not just something that
happens in like weird psych

501
00:20:10,160 --> 00:20:12,440
demos and funny movies.

502
00:20:12,440 --> 00:20:17,480
Even for very important things,
people miss absolutely major,

503
00:20:17,480 --> 00:20:19,600
important information.

504
00:20:19,600 --> 00:20:23,500
OK, so what all this
tells us is attention

505
00:20:23,500 --> 00:20:28,060
is a filter that lets some
information into our awareness

506
00:20:28,060 --> 00:20:30,760
but completely filters
out of awareness

507
00:20:30,760 --> 00:20:32,890
a lot of other
information that lands

508
00:20:32,890 --> 00:20:35,410
on your retina or your cochlea.

509
00:20:35,410 --> 00:20:37,660
OK?

510
00:20:37,660 --> 00:20:39,880
And there's two key
properties of attention

511
00:20:39,880 --> 00:20:41,110
that go hand-in-hand.

512
00:20:41,110 --> 00:20:43,360
One, those capacity limits
we've been talking about.

513
00:20:43,360 --> 00:20:45,290
You can't do everything at once.

514
00:20:45,290 --> 00:20:48,160
And two, selectivity,
that is there's

515
00:20:48,160 --> 00:20:50,470
some way some subset
of that information

516
00:20:50,470 --> 00:20:54,000
comes in some way
to select it, OK?

517
00:20:54,000 --> 00:20:57,030
All right, so let's say a
little bit more about attention.

518
00:20:57,030 --> 00:20:59,760
There's lots of different
ways and forms of attention.

519
00:20:59,760 --> 00:21:04,420
So let me give you a few key
distinctions about attention.

520
00:21:04,420 --> 00:21:06,420
So far, it's just kind
of a vague word that gets

521
00:21:06,420 --> 00:21:08,860
used lots of different ways.

522
00:21:08,860 --> 00:21:13,620
So long ago, the great Helmholtz
thought about attention

523
00:21:13,620 --> 00:21:16,650
and realized that there were two
different kinds of attention.

524
00:21:16,650 --> 00:21:20,880
So he noted way back
in 1860, our attention

525
00:21:20,880 --> 00:21:23,760
is quite independent of the
position and accommodation

526
00:21:23,760 --> 00:21:27,840
of the eyes and of any known
alteration in these organs

527
00:21:27,840 --> 00:21:31,860
and free to direct itself by a
conscious and voluntary effort

528
00:21:31,860 --> 00:21:35,400
upon any selected portion of a
dark and undifferentiated field

529
00:21:35,400 --> 00:21:36,460
of view.

530
00:21:36,460 --> 00:21:38,460
This is one of the most
important observations

531
00:21:38,460 --> 00:21:40,890
for a future theory
of attention.

532
00:21:40,890 --> 00:21:43,050
Indeed, it is, right?

533
00:21:43,050 --> 00:21:45,780
So his point-- I think we did
this at some earlier lecture.

534
00:21:45,780 --> 00:21:49,350
His point is you can fixate
on my nose right now.

535
00:21:49,350 --> 00:21:51,540
OK, everybody fixate on my nose.

536
00:21:51,540 --> 00:21:53,070
Not that it's that
fabulous a nose.

537
00:21:53,070 --> 00:21:55,020
It'll just serve for
the demo right now.

538
00:21:55,020 --> 00:21:58,860
And if I hold up different
numbers of fingers out here--

539
00:21:58,860 --> 00:22:00,120
keep fixating on my nose.

540
00:22:00,120 --> 00:22:01,530
No cheating.

541
00:22:01,530 --> 00:22:02,700
OK?

542
00:22:02,700 --> 00:22:06,570
How many fingers are off
to the left side of my nose

543
00:22:06,570 --> 00:22:08,280
from your perspective?

544
00:22:08,280 --> 00:22:09,060
OK.

545
00:22:09,060 --> 00:22:11,250
How many fingers are
off to the right side?

546
00:22:11,250 --> 00:22:12,060
OK.

547
00:22:12,060 --> 00:22:13,020
You can do that.

548
00:22:13,020 --> 00:22:15,750
You can pull up this information
or pull up that information

549
00:22:15,750 --> 00:22:17,640
without moving your eyes.

550
00:22:17,640 --> 00:22:18,180
OK?

551
00:22:18,180 --> 00:22:22,380
That's called covert attention
because the input is identical.

552
00:22:22,380 --> 00:22:24,240
And you are just
somehow adjusting

553
00:22:24,240 --> 00:22:25,800
the dials on your
perceptual system

554
00:22:25,800 --> 00:22:28,570
to pull up this or pull up that.

555
00:22:28,570 --> 00:22:29,080
OK?

556
00:22:29,080 --> 00:22:31,840
So that's called
covert attention.

557
00:22:31,840 --> 00:22:34,150
To be unconfounded
from overt attention,

558
00:22:34,150 --> 00:22:36,430
which is actually the most
powerful kind of attention.

559
00:22:36,430 --> 00:22:38,890
Overt attention, you
move your eyes from point

560
00:22:38,890 --> 00:22:41,830
to point to select
different information.

561
00:22:41,830 --> 00:22:46,420
OK, so overt attention is a
much more powerful filter.

562
00:22:46,420 --> 00:22:49,570
And we make about two to
four eye movements a second.

563
00:22:49,570 --> 00:22:52,000
And it's very powerful
because the center of gaze

564
00:22:52,000 --> 00:22:53,890
has very high
resolution information.

565
00:22:53,890 --> 00:22:55,840
Remember high density
of photoreceptors

566
00:22:55,840 --> 00:22:58,030
at the fovea in
the center of gaze,

567
00:22:58,030 --> 00:23:01,070
and the density tails
off in the periphery.

568
00:23:01,070 --> 00:23:04,060
So we have much better
information in the fovea

569
00:23:04,060 --> 00:23:06,040
than in the periphery.

570
00:23:06,040 --> 00:23:07,990
OK, so it's both
photoreceptor density,

571
00:23:07,990 --> 00:23:10,540
and it's also cortical area.

572
00:23:10,540 --> 00:23:12,910
So in retinotopic
cortex, back here,

573
00:23:12,910 --> 00:23:16,300
primary visual cortex,
V1, V2, those regions,

574
00:23:16,300 --> 00:23:20,260
you have about 20 square
centimeters of cortex,

575
00:23:20,260 --> 00:23:22,000
like that area of cortex--

576
00:23:22,000 --> 00:23:25,750
pretty big-- something like
that processing the central two

577
00:23:25,750 --> 00:23:27,350
degrees of vision.

578
00:23:27,350 --> 00:23:27,850
OK?

579
00:23:27,850 --> 00:23:30,820
Huge cortical area devoted
to a tiny little part

580
00:23:30,820 --> 00:23:32,120
of the visual world.

581
00:23:32,120 --> 00:23:36,340
And much less per degree
for the periphery.

582
00:23:36,340 --> 00:23:38,230
OK, so both at the
photoreceptor stage

583
00:23:38,230 --> 00:23:40,060
and at higher-level
processing stages,

584
00:23:40,060 --> 00:23:43,180
we vastly over represent
the center of gaze.

585
00:23:43,180 --> 00:23:43,930
OK?

586
00:23:43,930 --> 00:23:46,690
That's why if you have loss
of foveal vision and macular

587
00:23:46,690 --> 00:23:48,190
degeneration, it's really awful.

588
00:23:48,190 --> 00:23:49,990
You can't read or
recognize faces.

589
00:23:49,990 --> 00:23:52,000
That's where you need
this fine-grained foveal

590
00:23:52,000 --> 00:23:53,260
information.

591
00:23:53,260 --> 00:23:56,860
OK, so that means
that moving your fovea

592
00:23:56,860 --> 00:23:58,930
and parking it on different
parts of the array

593
00:23:58,930 --> 00:24:01,360
is an extremely
powerful way to select

594
00:24:01,360 --> 00:24:03,350
different kinds of information.

595
00:24:03,350 --> 00:24:06,340
So to give you a feel for
how much people do this

596
00:24:06,340 --> 00:24:11,020
and how they select
information, here's a video.

597
00:24:11,020 --> 00:24:12,820
This is taken from a
head-mounted camera

598
00:24:12,820 --> 00:24:15,310
of a person who is
watching these two people.

599
00:24:15,310 --> 00:24:18,130
And the yellow dot is
where their fovea is.

600
00:24:18,130 --> 00:24:20,470
So they're talking to
my former postdoc Matt.

601
00:24:20,470 --> 00:24:23,380
And they're fixating on
his face as he talks.

602
00:24:23,380 --> 00:24:26,530
And sometimes they take a
peek at the other person.

603
00:24:26,530 --> 00:24:29,380
And you can see there's very
fine-grained adjustments

604
00:24:29,380 --> 00:24:31,198
of where they look
as this goes on.

605
00:24:31,198 --> 00:24:33,490
And now they're walking down
the hall in this building,

606
00:24:33,490 --> 00:24:34,915
and they're following her.

607
00:24:34,915 --> 00:24:36,790
Watch what happens when
they turn the corner.

608
00:24:36,790 --> 00:24:38,110
Check out the signs.

609
00:24:38,110 --> 00:24:40,150
Read the little notices.

610
00:24:40,150 --> 00:24:41,240
Look down the hall.

611
00:24:41,240 --> 00:24:43,698
And then watch what happens
when they go around the corner.

612
00:24:43,698 --> 00:24:44,500
Oh, new person.

613
00:24:44,500 --> 00:24:46,010
Look at the new person.

614
00:24:46,010 --> 00:24:46,510
Right?

615
00:24:49,310 --> 00:24:51,410
Sorry, it's too dark to
see what's going on here.

616
00:24:51,410 --> 00:24:52,940
Oh, this is a little
fixation patch

617
00:24:52,940 --> 00:24:56,420
where they're recalibrating
the eye tracker to figure out

618
00:24:56,420 --> 00:24:57,700
to make sure it's working.

619
00:24:57,700 --> 00:24:58,700
Here comes a new person.

620
00:24:58,700 --> 00:25:00,110
They look at his face.

621
00:25:00,110 --> 00:25:03,890
Saccade back and forth
between the two faces, right?

622
00:25:03,890 --> 00:25:05,690
Whoever's talking,
you look at them.

623
00:25:05,690 --> 00:25:07,850
And you collect
high-resolution information

624
00:25:07,850 --> 00:25:08,958
from that person's face.

625
00:25:08,958 --> 00:25:11,000
With both head turns, you
can see their both head

626
00:25:11,000 --> 00:25:15,520
turns and eye movements.

627
00:25:15,520 --> 00:25:17,140
OK, you get the idea.

628
00:25:17,140 --> 00:25:18,250
I didn't totally explain.

629
00:25:18,250 --> 00:25:22,922
So this is a head-mounted
camera and eye tracker, OK,

630
00:25:22,922 --> 00:25:24,880
of a person who was
walking around the building

631
00:25:24,880 --> 00:25:25,770
encountering people.

632
00:25:25,770 --> 00:25:28,228
And it was showing you both
what was in their field of view

633
00:25:28,228 --> 00:25:30,890
and where they were fixating
their eyes in that view.

634
00:25:30,890 --> 00:25:33,700
So you can see there's really
fine-grained, moment-to-moment

635
00:25:33,700 --> 00:25:35,620
sampling of visual
information all

636
00:25:35,620 --> 00:25:38,800
the time with overt attention
with eye movements, OK?

637
00:25:38,800 --> 00:25:40,120
Make sense?

638
00:25:40,120 --> 00:25:41,082
OK.

639
00:25:41,082 --> 00:25:43,540
OK, first, so that's the first
distinction about attention,

640
00:25:43,540 --> 00:25:45,490
covert versus overt.

641
00:25:45,490 --> 00:25:46,780
OK?

642
00:25:46,780 --> 00:25:48,700
Now, you might say,
why would we bother

643
00:25:48,700 --> 00:25:52,120
with this subtle, sophisticated
covert attention when

644
00:25:52,120 --> 00:25:56,100
we can just move our eyes
and do overt attention?

645
00:25:56,100 --> 00:25:58,100
And I think there's a
bunch of reasons for that.

646
00:25:58,100 --> 00:26:01,720
One is other people can
see where you're looking.

647
00:26:01,720 --> 00:26:04,690
And there's sometimes you want
to attend to stuff over there

648
00:26:04,690 --> 00:26:07,390
and not let people know, right?

649
00:26:07,390 --> 00:26:11,620
So there are many examples of
this from elevator eyes, people

650
00:26:11,620 --> 00:26:13,202
who look you up and down.

651
00:26:13,202 --> 00:26:14,410
It's not politically correct.

652
00:26:14,410 --> 00:26:15,250
It's not nice.

653
00:26:15,250 --> 00:26:16,480
It's not considerate.

654
00:26:16,480 --> 00:26:18,830
People notice if you do that.

655
00:26:18,830 --> 00:26:19,600
So don't do it.

656
00:26:19,600 --> 00:26:23,080
You can covertly attend
below the face, if you like.

657
00:26:23,080 --> 00:26:25,450
But don't overtly move
your eyes down there

658
00:26:25,450 --> 00:26:26,830
because they will know.

659
00:26:26,830 --> 00:26:27,400
OK?

660
00:26:27,400 --> 00:26:29,380
We primates are
very, very attuned

661
00:26:29,380 --> 00:26:31,600
to where each other are looking.

662
00:26:31,600 --> 00:26:35,350
We are the only primate who
has whites around the eyes.

663
00:26:35,350 --> 00:26:37,660
I just heard a talk last
week by Michael Tomasello,

664
00:26:37,660 --> 00:26:40,000
who's one of the
major primatologists.

665
00:26:40,000 --> 00:26:42,250
And he says that
the fact that humans

666
00:26:42,250 --> 00:26:44,770
have whites of their eyes that
make it so easy to tell where

667
00:26:44,770 --> 00:26:47,470
they're looking must
mean that we mostly

668
00:26:47,470 --> 00:26:50,157
want to share information
with each other about where

669
00:26:50,157 --> 00:26:50,740
we're looking.

670
00:26:50,740 --> 00:26:54,400
We're a very cooperative,
communicative species.

671
00:26:54,400 --> 00:26:55,363
But not always.

672
00:26:55,363 --> 00:26:57,280
Sometimes we want to
sneak a peek at something

673
00:26:57,280 --> 00:26:58,510
where people don't know.

674
00:26:58,510 --> 00:27:00,160
The case that I
notice all the time is

675
00:27:00,160 --> 00:27:02,290
I'm at a conference and
some very familiar face

676
00:27:02,290 --> 00:27:03,940
comes along and
says, hey, Nancy.

677
00:27:03,940 --> 00:27:04,450
How are you?

678
00:27:04,450 --> 00:27:06,520
And I'm thinking,
who the hell is this?

679
00:27:06,520 --> 00:27:09,460
And the whole time, I'm thinking
I can see the name right there.

680
00:27:09,460 --> 00:27:10,600
I can't quite resolve it.

681
00:27:10,600 --> 00:27:13,480
I know that if saccade
down for a half a second,

682
00:27:13,480 --> 00:27:14,200
they will see it.

683
00:27:14,200 --> 00:27:15,790
And I will be busted.

684
00:27:15,790 --> 00:27:18,310
And so I'm sitting there with
a name right there racking

685
00:27:18,310 --> 00:27:20,290
my brain trying to think
who the hell it is.

686
00:27:20,290 --> 00:27:22,623
Anyway, so there are many
cases like this where we don't

687
00:27:22,623 --> 00:27:24,767
want to be caught looking.

688
00:27:24,767 --> 00:27:26,350
So those are all
cases where you might

689
00:27:26,350 --> 00:27:28,870
want to use covert attention.

690
00:27:28,870 --> 00:27:30,670
Also, sometimes
you want to track

691
00:27:30,670 --> 00:27:34,240
lots of things in parallel, and
you can't foveate all of them.

692
00:27:34,240 --> 00:27:36,500
You can only foveate one thing.

693
00:27:36,500 --> 00:27:38,230
So I don't know
anything about sports.

694
00:27:38,230 --> 00:27:40,480
But think of your favorite
complicated sports example

695
00:27:40,480 --> 00:27:42,438
where there are multiple
players moving at once

696
00:27:42,438 --> 00:27:44,270
and you need to keep
track of all of them.

697
00:27:44,270 --> 00:27:46,270
That would be a classic
case of covert attention

698
00:27:46,270 --> 00:27:48,400
because you could covertly
attend to several of them

699
00:27:48,400 --> 00:27:48,900
at once.

700
00:27:48,900 --> 00:27:50,720
And you have only one fovea.

701
00:27:50,720 --> 00:27:53,860
Well, you have two, but
they go to the same place.

702
00:27:53,860 --> 00:27:55,720
OK, so that's the
second distinction overt

703
00:27:55,720 --> 00:27:58,150
versus covert--

704
00:27:58,150 --> 00:27:58,930
or the first one.

705
00:27:58,930 --> 00:27:59,430
Right?

706
00:27:59,430 --> 00:28:03,100
So the next basic distinction
of different kinds of attention

707
00:28:03,100 --> 00:28:05,350
are the kind where
we decide where

708
00:28:05,350 --> 00:28:09,400
we want to look versus the
kind where the stimulus draws

709
00:28:09,400 --> 00:28:11,230
our attention, OK?

710
00:28:11,230 --> 00:28:13,510
So there are lots of examples.

711
00:28:13,510 --> 00:28:15,880
Like web pop ups
are all designed

712
00:28:15,880 --> 00:28:18,550
to pull your attention,
overt and covert.

713
00:28:18,550 --> 00:28:20,710
Even you don't want to
look at that damn thing.

714
00:28:20,710 --> 00:28:22,870
But the weird, little
dancing figure,

715
00:28:22,870 --> 00:28:26,710
they're all optimized to pull
your attention over and make

716
00:28:26,710 --> 00:28:28,652
you read the ad, right?

717
00:28:28,652 --> 00:28:29,860
And they're pretty effective.

718
00:28:29,860 --> 00:28:32,860
OK, so that's called
stimulus-driven attention

719
00:28:32,860 --> 00:28:34,027
or exogenous attention.

720
00:28:34,027 --> 00:28:35,110
It comes from the outside.

721
00:28:35,110 --> 00:28:36,670
And it pulls you in.

722
00:28:36,670 --> 00:28:38,950
OK?

723
00:28:38,950 --> 00:28:41,080
Another classic
example is pop-out.

724
00:28:41,080 --> 00:28:42,880
If you look at this
display, it's very hard

725
00:28:42,880 --> 00:28:44,710
not to notice and have
your attention drawn

726
00:28:44,710 --> 00:28:46,450
to that red thing.

727
00:28:46,450 --> 00:28:48,340
Certain properties
of stimuli that just

728
00:28:48,340 --> 00:28:50,830
automatically capture
your attention.

729
00:28:50,830 --> 00:28:53,200
OK, that is to be
contrasted with

730
00:28:53,200 --> 00:28:54,713
voluntary controlled attention.

731
00:28:54,713 --> 00:28:56,380
That's like the case
we did before where

732
00:28:56,380 --> 00:28:58,030
you were fixating on my nose.

733
00:28:58,030 --> 00:29:00,083
It was totally up to
you whether attend--

734
00:29:00,083 --> 00:29:01,000
I mean, I told you to.

735
00:29:01,000 --> 00:29:02,350
But you could have
made up your own mind

736
00:29:02,350 --> 00:29:03,460
and attended to neither.

737
00:29:03,460 --> 00:29:05,350
And I wouldn't
have known, right?

738
00:29:05,350 --> 00:29:07,910
So you can decide what you
want to pay attention to.

739
00:29:07,910 --> 00:29:09,520
And as I think I
mentioned in I don't

740
00:29:09,520 --> 00:29:12,040
know what context
a few weeks ago,

741
00:29:12,040 --> 00:29:14,188
thank God we can decide
what to pay attention to.

742
00:29:14,188 --> 00:29:16,480
Because there's lots of stuff
that goes on in the world

743
00:29:16,480 --> 00:29:18,970
that we don't want to have
dominate our mental life.

744
00:29:18,970 --> 00:29:22,390
And controlled or
voluntary attention

745
00:29:22,390 --> 00:29:25,930
is one way that we can have
our mental life dominated

746
00:29:25,930 --> 00:29:28,420
more by things we want it
to be dominated by and less

747
00:29:28,420 --> 00:29:32,170
by things we don't, right?

748
00:29:32,170 --> 00:29:33,790
So that's that notion here.

749
00:29:33,790 --> 00:29:36,190
Like for example,
you're sitting here.

750
00:29:36,190 --> 00:29:37,180
And it's like noon.

751
00:29:37,180 --> 00:29:38,860
And you're just thinking,
OK, I'm really hungry.

752
00:29:38,860 --> 00:29:39,730
I'm really hungry.

753
00:29:39,730 --> 00:29:42,130
If you were my dog Charlie,
your entire mental life

754
00:29:42,130 --> 00:29:44,080
for the whole next half an
hour would be, I'm so hungry.

755
00:29:44,080 --> 00:29:44,663
I'm so hungry.

756
00:29:44,663 --> 00:29:45,760
I'm so hungry.

757
00:29:45,760 --> 00:29:47,000
But you're a person.

758
00:29:47,000 --> 00:29:49,235
And so that thought may
impinge now and then,

759
00:29:49,235 --> 00:29:51,610
but you can drive it away and
think about something else.

760
00:29:51,610 --> 00:29:53,620
Because it's not going
to get you anywhere

761
00:29:53,620 --> 00:29:54,913
to focus on how hungry you are.

762
00:29:54,913 --> 00:29:56,080
I know I just made it worse.

763
00:29:56,080 --> 00:29:59,170
I'm sorry about that.

764
00:29:59,170 --> 00:30:00,840
OK.

765
00:30:00,840 --> 00:30:04,750
OK, the way that
scientists have typically

766
00:30:04,750 --> 00:30:08,830
studied voluntary attention is
they ask subjects to do this.

767
00:30:08,830 --> 00:30:10,300
And much like the
case we just did,

768
00:30:10,300 --> 00:30:12,580
here's kind of very
basic paradigm.

769
00:30:12,580 --> 00:30:14,380
You have subjects
fixating on a cross.

770
00:30:14,380 --> 00:30:17,410
And typically, in covert
attention experiments,

771
00:30:17,410 --> 00:30:19,960
you want to make sure there
aren't overt eye movements.

772
00:30:19,960 --> 00:30:21,460
And so you'll have an
eye tracker to make sure

773
00:30:21,460 --> 00:30:23,470
the subject is keeping
their eye on that cross.

774
00:30:23,470 --> 00:30:24,472
OK?

775
00:30:24,472 --> 00:30:25,930
And then you give
them a little cue

776
00:30:25,930 --> 00:30:28,030
that says pay attention
off to the right,

777
00:30:28,030 --> 00:30:30,700
but don't move your eyes, OK?

778
00:30:30,700 --> 00:30:33,730
And then shortly thereafter,
a little cue comes up.

779
00:30:33,730 --> 00:30:35,230
And you have to hit
a button quickly

780
00:30:35,230 --> 00:30:36,580
that you detected the target.

781
00:30:36,580 --> 00:30:36,850
OK?

782
00:30:36,850 --> 00:30:38,183
It comes out at variable delays.

783
00:30:38,183 --> 00:30:40,450
So you can't just
hit the button.

784
00:30:40,450 --> 00:30:43,060
OK, if you do that, you
find that the reaction

785
00:30:43,060 --> 00:30:45,760
time to detect that target, if
it's where you're attending,

786
00:30:45,760 --> 00:30:48,860
is really fast, not much
more than 200 milliseconds,

787
00:30:48,860 --> 00:30:51,520
which is damned fast, right?

788
00:30:51,520 --> 00:30:54,765
Now consider a case where you're
instructed to attend over here.

789
00:30:54,765 --> 00:30:56,640
And the way you get
these experiments to work

790
00:30:56,640 --> 00:30:59,167
is that these cues are
valid 80% of the time.

791
00:30:59,167 --> 00:31:01,000
So there's a reason for
subject to-- there's

792
00:31:01,000 --> 00:31:03,120
a motivation for them to
follow your instruction.

793
00:31:03,120 --> 00:31:04,380
So they're trying
to answer quickly.

794
00:31:04,380 --> 00:31:05,340
And there's like,
OK, they're ready.

795
00:31:05,340 --> 00:31:06,690
It's almost certainly
going to be over here.

796
00:31:06,690 --> 00:31:07,260
But oops.

797
00:31:07,260 --> 00:31:08,160
It's not.

798
00:31:08,160 --> 00:31:09,180
OK?

799
00:31:09,180 --> 00:31:12,600
Then reaction time is almost
100 milliseconds slower,

800
00:31:12,600 --> 00:31:16,350
which is a huge effect in
psychology, 100 milliseconds.

801
00:31:16,350 --> 00:31:17,310
OK?

802
00:31:17,310 --> 00:31:20,880
OK, so you're much slower
to detect something

803
00:31:20,880 --> 00:31:23,320
if it's at an
unattended location.

804
00:31:23,320 --> 00:31:23,820
OK?

805
00:31:23,820 --> 00:31:26,910
And if you have a neutral trial,
you're somewhere in the middle.

806
00:31:26,910 --> 00:31:33,120
OK, so that is exogenous
or voluntary attention.

807
00:31:33,120 --> 00:31:35,340
OK, and again, this
is covert attention.

808
00:31:35,340 --> 00:31:37,410
No eye movements.

809
00:31:37,410 --> 00:31:41,460
All right, so third
distinction, the example

810
00:31:41,460 --> 00:31:43,230
I just gave you was
spatial attention.

811
00:31:43,230 --> 00:31:46,140
You're attending to this
location or that location

812
00:31:46,140 --> 00:31:47,410
in space.

813
00:31:47,410 --> 00:31:47,910
OK?

814
00:31:47,910 --> 00:31:50,340
That's probably the most
powerful and common kind

815
00:31:50,340 --> 00:31:52,432
of attention is
usually the things

816
00:31:52,432 --> 00:31:54,390
you're interested in are
in a particular place.

817
00:31:54,390 --> 00:31:55,598
And you attend to that place.

818
00:31:55,598 --> 00:31:57,360
And there you go.

819
00:31:57,360 --> 00:31:59,130
But it's not the only kind.

820
00:31:59,130 --> 00:32:01,360
You can have
feature-based attention.

821
00:32:01,360 --> 00:32:04,680
So if your utensil drawer is
a hell of a mess like mine

822
00:32:04,680 --> 00:32:09,540
is, like this, if your task
is to find the black vegetable

823
00:32:09,540 --> 00:32:12,030
peeler, best of luck to you.

824
00:32:12,030 --> 00:32:13,710
It'll take quite a while.

825
00:32:13,710 --> 00:32:14,380
It's in there.

826
00:32:14,380 --> 00:32:15,088
It's right there.

827
00:32:15,088 --> 00:32:16,800
But it takes a while, right?

828
00:32:16,800 --> 00:32:21,660
In contrast, if the task is find
the purple rubber spatula, duh,

829
00:32:21,660 --> 00:32:22,440
right?

830
00:32:22,440 --> 00:32:24,180
OK, so that's
feature-based attention.

831
00:32:24,180 --> 00:32:25,950
You don't know the
location in advance.

832
00:32:25,950 --> 00:32:28,283
You know something about the
feature you're looking for,

833
00:32:28,283 --> 00:32:31,140
that it's black or that it's
purple or round or square

834
00:32:31,140 --> 00:32:32,610
or whatever the feature is.

835
00:32:32,610 --> 00:32:35,070
OK, so different
kinds of filters

836
00:32:35,070 --> 00:32:38,040
you can set on your
attention system.

837
00:32:38,040 --> 00:32:39,810
OK, so that's just
to give you some

838
00:32:39,810 --> 00:32:42,640
of the phenomenology of the
different kinds of attention.

839
00:32:42,640 --> 00:32:44,245
How does all this
work in the brain?

840
00:32:44,245 --> 00:32:45,870
I'm not going to tell
you how it works.

841
00:32:45,870 --> 00:32:49,920
We'll just focus mostly on
what the brain regions are.

842
00:32:49,920 --> 00:32:55,750
OK, so let's go back to this
case here, voluntary attention.

843
00:32:55,750 --> 00:32:59,160
So now the question is, what
happens when you get this cue?

844
00:32:59,160 --> 00:33:00,690
You're not going
to move your eyes.

845
00:33:00,690 --> 00:33:03,900
But you're kind of cranking
up that part of space.

846
00:33:03,900 --> 00:33:07,020
OK, what's going on in
the brain right there

847
00:33:07,020 --> 00:33:09,150
before the stimulus comes on?

848
00:33:09,150 --> 00:33:10,860
Everybody get the question?

849
00:33:10,860 --> 00:33:12,900
OK, you can sort of
feel it happening.

850
00:33:12,900 --> 00:33:14,830
But what does that mean?

851
00:33:14,830 --> 00:33:18,090
Well, this is very amenable
to a simple functional MRI

852
00:33:18,090 --> 00:33:19,110
experiment.

853
00:33:19,110 --> 00:33:21,360
And in the early days
of functional MRI,

854
00:33:21,360 --> 00:33:24,210
a whole bunch of people at
once realize, oh, right, we

855
00:33:24,210 --> 00:33:26,190
can answer this
longstanding question

856
00:33:26,190 --> 00:33:29,640
of whether early retinotopic
parts of the visual system

857
00:33:29,640 --> 00:33:32,610
are affected just
by a cue like that,

858
00:33:32,610 --> 00:33:36,150
even before the stimulus
comes on, right?

859
00:33:36,150 --> 00:33:38,430
And I think it was 1998.

860
00:33:38,430 --> 00:33:40,980
A whole bunch of people
started doing this in 1997.

861
00:33:40,980 --> 00:33:42,420
I was one of them.

862
00:33:42,420 --> 00:33:44,340
But a whole bunch
did it once in 1998.

863
00:33:44,340 --> 00:33:46,647
I think like six
papers were published

864
00:33:46,647 --> 00:33:48,480
all doing versions of
this next thing, which

865
00:33:48,480 --> 00:33:51,150
is one of those obvious ideas.

866
00:33:51,150 --> 00:33:51,900
We got scooped.

867
00:33:51,900 --> 00:33:54,730
But anyway, everybody
got the same result.

868
00:33:54,730 --> 00:33:57,070
So basically, if you just
do a simple comparison--

869
00:33:57,070 --> 00:33:58,320
this is the version they have.

870
00:33:58,320 --> 00:33:59,520
Subjects are fixating here.

871
00:33:59,520 --> 00:34:01,410
There's two stimuli.

872
00:34:01,410 --> 00:34:03,420
They're looking in V1 and V2.

873
00:34:03,420 --> 00:34:06,180
And if you just do a contrast
of the case where you're

874
00:34:06,180 --> 00:34:08,639
looking here and you're
attending, waiting for a target

875
00:34:08,639 --> 00:34:12,510
here versus waiting for a target
there without moving your eyes,

876
00:34:12,510 --> 00:34:17,429
you get a big contralateral
modulation in V1 and V2.

877
00:34:17,429 --> 00:34:19,750
This is a slice near the
back of the head here.

878
00:34:19,750 --> 00:34:22,260
So this is a piece
of calcarine sulcus

879
00:34:22,260 --> 00:34:24,929
or primary visual cortex, right?

880
00:34:24,929 --> 00:34:28,620
And so you can see, even before
the target stimulus comes on,

881
00:34:28,620 --> 00:34:32,639
just attending over there gives
you a big baseline increase

882
00:34:32,639 --> 00:34:33,570
in neural activity.

883
00:34:33,570 --> 00:34:35,580
Everybody clear
what's going on here?

884
00:34:35,580 --> 00:34:37,480
And of course, you can
do the opposite one

885
00:34:37,480 --> 00:34:40,409
and see that it shifts to
the opposite side of space.

886
00:34:40,409 --> 00:34:41,750
OK?

887
00:34:41,750 --> 00:34:44,600
OK, so that's all before
the stimulus comes on.

888
00:34:44,600 --> 00:34:47,360
You can think of it as priming
the brain, kind of juicing up

889
00:34:47,360 --> 00:34:49,020
those neurons getting
them ready to go,

890
00:34:49,020 --> 00:34:51,020
so that, when that stimulus
comes through, boom.

891
00:34:51,020 --> 00:34:53,199
Your reaction time is faster.

892
00:34:53,199 --> 00:34:55,570
OK?

893
00:34:55,570 --> 00:34:59,920
OK, so that's modulating
parts of visual cortex

894
00:34:59,920 --> 00:35:03,100
before a stimulus comes
on to get it ready.

895
00:35:03,100 --> 00:35:05,020
But we can also look
at, what is the effect

896
00:35:05,020 --> 00:35:07,000
on a stimulus once
it comes through,

897
00:35:07,000 --> 00:35:09,665
an attended stimulus or
an unattended stimulus?

898
00:35:09,665 --> 00:35:11,290
To show you that,
I'm going to show you

899
00:35:11,290 --> 00:35:16,360
a kind of feature-based
attention and an ancient-- this

900
00:35:16,360 --> 00:35:19,540
is a paper we published
in 1998, I think.

901
00:35:19,540 --> 00:35:21,460
This is a nonspatial version.

902
00:35:21,460 --> 00:35:23,950
So we gave subjects--
guess what-- faces

903
00:35:23,950 --> 00:35:26,260
and houses and a
little crosshair.

904
00:35:26,260 --> 00:35:28,750
And we had subjects
in different blocks

905
00:35:28,750 --> 00:35:31,120
say whether the two
arms of the cross

906
00:35:31,120 --> 00:35:34,330
were the same or
different length, OK?

907
00:35:34,330 --> 00:35:35,855
So vertical one is
different length.

908
00:35:35,855 --> 00:35:37,180
So that's a different case.

909
00:35:37,180 --> 00:35:40,420
Whether the two houses
are same or different

910
00:35:40,420 --> 00:35:42,770
or whether the two faces
are same or different.

911
00:35:42,770 --> 00:35:44,680
OK, so actually, it's
spatial and feature,

912
00:35:44,680 --> 00:35:47,245
when you think of it, because
they're in different places.

913
00:35:47,245 --> 00:35:48,100
OK?

914
00:35:48,100 --> 00:35:50,350
So we then looked in
a bunch of regions.

915
00:35:50,350 --> 00:35:53,140
I'll show you the data from
the fusiform face area.

916
00:35:53,140 --> 00:35:56,740
And the key thing about
this is, in all three tasks,

917
00:35:56,740 --> 00:35:59,300
we had the identical stimuli.

918
00:35:59,300 --> 00:36:01,570
So there's always
faces and houses

919
00:36:01,570 --> 00:36:04,330
and crosses in the same
place in your visual field,

920
00:36:04,330 --> 00:36:05,830
in all of those blocks.

921
00:36:05,830 --> 00:36:07,690
It's just a matter of
which of those things

922
00:36:07,690 --> 00:36:09,700
you're paying attention to.

923
00:36:09,700 --> 00:36:10,605
OK?

924
00:36:10,605 --> 00:36:12,730
So what do you think we'll
see in the fusiform face

925
00:36:12,730 --> 00:36:16,240
area in this experiment?

926
00:36:16,240 --> 00:36:20,170
Is it going to care whether
you're attending to faces

927
00:36:20,170 --> 00:36:23,590
or attending to houses or
attending to the crosshair?

928
00:36:23,590 --> 00:36:25,195
Same stimulus all the time.

929
00:36:29,840 --> 00:36:31,200
Yeah, Jimmy?

930
00:36:31,200 --> 00:36:33,200
AUDIENCE: [INAUDIBLE]
when you're not attending.

931
00:36:33,200 --> 00:36:36,040
But when you're
attending, [INAUDIBLE]

932
00:36:36,040 --> 00:36:37,863
NANCY KANWISHER:
Exactly what happens.

933
00:36:37,863 --> 00:36:39,280
OK, everybody hear
the prediction?

934
00:36:39,280 --> 00:36:40,780
You'll still get something.

935
00:36:40,780 --> 00:36:43,840
But you'll get more when
you're paying attention to it.

936
00:36:43,840 --> 00:36:47,690
OK, so here's the FFA
response over time.

937
00:36:47,690 --> 00:36:49,120
Here's the time course.

938
00:36:49,120 --> 00:36:51,400
And these gray blocks
are the blocks.

939
00:36:51,400 --> 00:36:53,560
These are the different
attending to houses, faces,

940
00:36:53,560 --> 00:36:57,370
color, houses, faces, crosshair,
houses, faces, crosshair.

941
00:36:57,370 --> 00:36:59,320
The gray bars are
when the subjects

942
00:36:59,320 --> 00:37:02,170
are attending to faces.

943
00:37:02,170 --> 00:37:09,020
Now, here, I actually
can't quite tell.

944
00:37:09,020 --> 00:37:10,480
Yeah, I can sort of tell.

945
00:37:10,480 --> 00:37:11,830
You can't see it on the screen.

946
00:37:11,830 --> 00:37:14,218
But there are little rest
periods between those blocks.

947
00:37:14,218 --> 00:37:15,760
So I'm trying to
figure out if you're

948
00:37:15,760 --> 00:37:19,630
right that there's some
response greater than baseline.

949
00:37:19,630 --> 00:37:21,970
I think there's actually very
little selective response

950
00:37:21,970 --> 00:37:23,290
greater than baseline.

951
00:37:23,290 --> 00:37:26,500
By design, we made this
a whopping attentional

952
00:37:26,500 --> 00:37:29,943
manipulation, so that all of
the tasks are really difficult.

953
00:37:29,943 --> 00:37:31,360
And when you're
doing one of them,

954
00:37:31,360 --> 00:37:33,680
you just barely feel
aware of anything else.

955
00:37:33,680 --> 00:37:36,100
In fact, I vividly
remember the first time

956
00:37:36,100 --> 00:37:38,710
we ran this experiment, I
was a subject in the scanner.

957
00:37:38,710 --> 00:37:40,720
And I did the first block.

958
00:37:40,720 --> 00:37:43,900
And I was doing just crosshairs.

959
00:37:43,900 --> 00:37:46,437
And I went through the
whole crosshair thing.

960
00:37:46,437 --> 00:37:48,770
And it wasn't till the end
of the experiment I realized,

961
00:37:48,770 --> 00:37:50,187
oh, right, there
were faces there.

962
00:37:50,187 --> 00:37:51,860
I was just completely
unaware of them.

963
00:37:51,860 --> 00:37:56,770
So we designed it to really
so tie up your mental capacity

964
00:37:56,770 --> 00:37:59,230
that you just didn't have
processing resources left

965
00:37:59,230 --> 00:38:00,200
for anything else.

966
00:38:00,200 --> 00:38:02,980
And so if there is a response
to everything else here,

967
00:38:02,980 --> 00:38:03,890
it's really low.

968
00:38:03,890 --> 00:38:06,670
But you can see, certainly, the
attended thing is much more.

969
00:38:06,670 --> 00:38:08,360
It's much stronger
in the attended case

970
00:38:08,360 --> 00:38:09,880
than the unattended case.

971
00:38:09,880 --> 00:38:10,720
OK?

972
00:38:10,720 --> 00:38:12,400
So this is a general property.

973
00:38:12,400 --> 00:38:14,440
Like basically, all of
the perceptual regions

974
00:38:14,440 --> 00:38:18,490
we've talked about are strongly
modulable by attention,

975
00:38:18,490 --> 00:38:23,320
OK, from retinotopic
regions on up.

976
00:38:23,320 --> 00:38:29,300
OK, including V1 and, in fact,
even including the lateral

977
00:38:29,300 --> 00:38:31,680
geniculate nucleus.

978
00:38:31,680 --> 00:38:34,520
So one synapse up
from the retina,

979
00:38:34,520 --> 00:38:38,390
you're already modulating
activity there by attention.

980
00:38:38,390 --> 00:38:40,880
OK?

981
00:38:40,880 --> 00:38:43,670
I know I said this, but
it maybe went by briefly.

982
00:38:43,670 --> 00:38:45,950
You have 10 times as
many connections down

983
00:38:45,950 --> 00:38:50,190
from cortex down to the LGN
as you have going forward.

984
00:38:50,190 --> 00:38:50,690
OK?

985
00:38:50,690 --> 00:38:52,790
One of the things they're
doing is setting up

986
00:38:52,790 --> 00:38:56,660
selective filters, so that only
the stuff you want to process

987
00:38:56,660 --> 00:38:59,250
makes it to higher stages.

988
00:38:59,250 --> 00:39:01,312
OK, Yes?

989
00:39:01,312 --> 00:39:03,020
AUDIENCE: So does this
sort of phenomenon

990
00:39:03,020 --> 00:39:05,230
generalize across
the brain areas?

991
00:39:05,230 --> 00:39:06,170
I mean, it's like--

992
00:39:06,170 --> 00:39:08,295
NANCY KANWISHER: It's
generally true of pretty much

993
00:39:08,295 --> 00:39:09,485
everywhere, yeah.

994
00:39:09,485 --> 00:39:11,280
AUDIENCE: It might be
the intuitive physics

995
00:39:11,280 --> 00:39:12,580
thing is justified.

996
00:39:12,580 --> 00:39:15,608
So it's better to see
the same stimulus.

997
00:39:15,608 --> 00:39:16,650
NANCY KANWISHER: Exactly.

998
00:39:16,650 --> 00:39:18,810
It's an instance of
this, exactly right.

999
00:39:18,810 --> 00:39:19,560
Exactly right.

1000
00:39:19,560 --> 00:39:22,110
It gives us one
of our paradigms.

1001
00:39:22,110 --> 00:39:24,660
In functional MRI studies,
have the same stimulus,

1002
00:39:24,660 --> 00:39:26,050
vary the task.

1003
00:39:26,050 --> 00:39:27,128
OK?

1004
00:39:27,128 --> 00:39:28,920
Some things don't work
as well, things that

1005
00:39:28,920 --> 00:39:31,030
are just very dominant stimuli.

1006
00:39:31,030 --> 00:39:33,900
In this experiment, if we hadn't
put the faces in the periphery

1007
00:39:33,900 --> 00:39:36,900
and given subjects a
damn difficult task,

1008
00:39:36,900 --> 00:39:38,400
the modulation
wouldn't have worked.

1009
00:39:38,400 --> 00:39:39,900
Because faces are
just so dominant,

1010
00:39:39,900 --> 00:39:42,990
they're going to punch through
even if you don't want to, OK?

1011
00:39:42,990 --> 00:39:44,597
Like web pop ups. You also--

1012
00:39:44,597 --> 00:39:46,680
I'm going to skip this
because we won't have time.

1013
00:39:46,680 --> 00:39:48,180
You could get very
similar things

1014
00:39:48,180 --> 00:39:51,000
in primary auditory cortex
listening to high frequencies

1015
00:39:51,000 --> 00:39:52,140
or low frequencies.

1016
00:39:52,140 --> 00:39:53,800
These are all the same stimuli.

1017
00:39:53,800 --> 00:39:56,860
This is voxel selected
for low frequencies, voxel

1018
00:39:56,860 --> 00:39:58,110
selected for high frequencies.

1019
00:39:58,110 --> 00:39:59,940
You have a high frequency
input in one ear,

1020
00:39:59,940 --> 00:40:01,080
low frequency in the other.

1021
00:40:01,080 --> 00:40:05,210
As you switch between,
those responses toggle.

1022
00:40:05,210 --> 00:40:08,390
OK, so all of that tells you
the effects of attention,

1023
00:40:08,390 --> 00:40:11,010
how it modulates activity
all over the brain.

1024
00:40:11,010 --> 00:40:14,510
But what is the source of
attention signals in the brain?

1025
00:40:14,510 --> 00:40:16,370
And that source is
a set of regions

1026
00:40:16,370 --> 00:40:19,130
we have encountered
before, sometimes called

1027
00:40:19,130 --> 00:40:21,050
the frontal-parietal
attention network,

1028
00:40:21,050 --> 00:40:23,180
these blue and green
bits up in here,

1029
00:40:23,180 --> 00:40:26,840
back here in the parietal lobe,
up here in the frontal lobe.

1030
00:40:26,840 --> 00:40:29,360
And they are active
when you shift attention

1031
00:40:29,360 --> 00:40:32,420
from one location to another,
from one feature to another,

1032
00:40:32,420 --> 00:40:36,260
or when you do any difficult
attention-demanding task.

1033
00:40:36,260 --> 00:40:38,660
We've also talked about
them in the other context

1034
00:40:38,660 --> 00:40:40,070
of the multiple demand system.

1035
00:40:40,070 --> 00:40:42,410
It's pretty much the same
set of cortical regions.

1036
00:40:42,410 --> 00:40:44,840
That pretty much any time
you do a difficult task,

1037
00:40:44,840 --> 00:40:48,240
almost no matter what the task
is, these regions turn on.

1038
00:40:48,240 --> 00:40:50,360
So they're not just
about visual attention.

1039
00:40:50,360 --> 00:40:54,890
They're kind of about basically
controlling your mind, right,

1040
00:40:54,890 --> 00:40:59,810
selecting information, making
yourself do difficult things.

1041
00:40:59,810 --> 00:41:01,910
Who knows what that
is computationally.

1042
00:41:01,910 --> 00:41:05,330
But these regions are very
systematically engaged.

1043
00:41:05,330 --> 00:41:07,550
And they're particularly
systematically engaged

1044
00:41:07,550 --> 00:41:10,970
when you're shifting attention
over different locations.

1045
00:41:10,970 --> 00:41:13,280
OK, so just to remind you
in contrast, everything else

1046
00:41:13,280 --> 00:41:16,610
we've talked about, face areas,
music areas, language areas,

1047
00:41:16,610 --> 00:41:18,980
they're very specific
for one mental process,

1048
00:41:18,980 --> 00:41:19,940
domain specific.

1049
00:41:19,940 --> 00:41:21,000
These are the opposite.

1050
00:41:21,000 --> 00:41:23,690
They're ludicrously
domain general.

1051
00:41:23,690 --> 00:41:25,280
OK?

1052
00:41:25,280 --> 00:41:28,580
All right, I'm going
to skip the video.

1053
00:41:28,580 --> 00:41:30,500
When the system is
damaged bilaterally,

1054
00:41:30,500 --> 00:41:33,360
all kinds of weird
stuff happens.

1055
00:41:33,360 --> 00:41:33,860
OK?

1056
00:41:33,860 --> 00:41:34,580
I'll post this.

1057
00:41:34,580 --> 00:41:36,500
If you guys send me
an email to remind me,

1058
00:41:36,500 --> 00:41:37,610
I'll post this clip in.

1059
00:41:37,610 --> 00:41:40,550
Balint's syndrome, where
you have bilateral damage

1060
00:41:40,550 --> 00:41:42,440
back there in these
parietal regions,

1061
00:41:42,440 --> 00:41:45,390
people can only see
one object at a time.

1062
00:41:45,390 --> 00:41:45,890
OK?

1063
00:41:45,890 --> 00:41:48,230
They're looking at a
complicated thing like this.

1064
00:41:48,230 --> 00:41:50,450
And they would see, I see Shosh.

1065
00:41:50,450 --> 00:41:51,320
See anything else?

1066
00:41:51,320 --> 00:41:52,520
No just Shosh.

1067
00:41:52,520 --> 00:41:53,390
Nothing else there?

1068
00:41:53,390 --> 00:41:53,890
No.

1069
00:41:53,890 --> 00:41:54,680
Just Shosh, right?

1070
00:41:54,680 --> 00:41:55,340
Totally weird.

1071
00:41:55,340 --> 00:41:57,170
Anyway, so they get
locked on one thing.

1072
00:41:57,170 --> 00:41:59,030
They can't shift attention.

1073
00:41:59,030 --> 00:42:01,550
OK.

1074
00:42:01,550 --> 00:42:05,220
OK, I want to talk at least
briefly about awareness.

1075
00:42:05,220 --> 00:42:09,780
So let's talk a little bit about
neural correlates of awareness.

1076
00:42:09,780 --> 00:42:13,220
And so first question
is, if we want

1077
00:42:13,220 --> 00:42:17,990
to study perceptual awareness
and not just perception,

1078
00:42:17,990 --> 00:42:20,570
how are we going to
uncouple those things?

1079
00:42:20,570 --> 00:42:23,720
We've sort of talked about a
few examples with attention.

1080
00:42:23,720 --> 00:42:25,940
With attention, the
thing you're attending to

1081
00:42:25,940 --> 00:42:28,220
is at least much more
dominant in your awareness

1082
00:42:28,220 --> 00:42:30,080
than the stuff you're
not attending to,

1083
00:42:30,080 --> 00:42:33,300
even if the other stuff
seeps in a little bit.

1084
00:42:33,300 --> 00:42:39,470
So that's one way, where you
have the identical stimulus.

1085
00:42:39,470 --> 00:42:42,140
But you were varying
the degree of awareness.

1086
00:42:42,140 --> 00:42:45,737
But another way-- oh, I need you
guys to pass out the glasses.

1087
00:42:45,737 --> 00:42:47,570
You want to both do
that at different sides?

1088
00:42:47,570 --> 00:42:49,517
OK, so I'm going to
show you a stimulus.

1089
00:42:49,517 --> 00:42:51,350
These guys are going
to pass around glasses.

1090
00:42:51,350 --> 00:42:52,940
And let me say in
advance, I want them back.

1091
00:42:52,940 --> 00:42:53,900
I use these every year.

1092
00:42:53,900 --> 00:42:55,822
So drop them off
here when you leave.

1093
00:42:55,822 --> 00:42:58,280
OK, so what you're going to
do, you put them on either way.

1094
00:42:58,280 --> 00:42:59,750
Doesn't matter.

1095
00:42:59,750 --> 00:43:01,700
OK, so first, we're
going to do optics.

1096
00:43:01,700 --> 00:43:03,260
Nothing interesting.

1097
00:43:03,260 --> 00:43:07,430
So look at this stimulus
and close one eye.

1098
00:43:07,430 --> 00:43:09,410
OK, now look at it
through the glasses.

1099
00:43:09,410 --> 00:43:11,690
Look at it through the
glasses, and close one eye.

1100
00:43:11,690 --> 00:43:14,630
And you should see just
a face or just a house.

1101
00:43:14,630 --> 00:43:17,400
And if you close the other eye,
you should see the opposite.

1102
00:43:17,400 --> 00:43:18,680
Does everybody get that?

1103
00:43:18,680 --> 00:43:19,910
OK, that's not psychology.

1104
00:43:19,910 --> 00:43:20,750
That's optics.

1105
00:43:20,750 --> 00:43:22,730
The glasses are just
filtering one image

1106
00:43:22,730 --> 00:43:26,385
into one eye and another
image into the other eye, OK?

1107
00:43:26,385 --> 00:43:28,760
This is all this is is a way
to get different information

1108
00:43:28,760 --> 00:43:30,110
to your two eyes.

1109
00:43:30,110 --> 00:43:32,690
OK, now just look
with both eyes.

1110
00:43:32,690 --> 00:43:33,890
And don't do anything.

1111
00:43:33,890 --> 00:43:35,270
Just kind of look.

1112
00:43:35,270 --> 00:43:39,360
And if anything cool happens,
you can kind of go, ooh, aah.

1113
00:43:43,480 --> 00:43:45,160
Or you can tell me
what's happening.

1114
00:43:45,160 --> 00:43:45,910
Just watch.

1115
00:43:45,910 --> 00:43:47,500
Doesn't always
happen immediately.

1116
00:43:47,500 --> 00:43:49,774
Yeah, Kwylie, what's happening?

1117
00:43:49,774 --> 00:43:51,160
AUDIENCE: [INAUDIBLE]

1118
00:43:51,160 --> 00:43:51,490
NANCY KANWISHER: Sorry.

1119
00:43:51,490 --> 00:43:51,990
What's that?

1120
00:43:51,990 --> 00:43:54,370
AUDIENCE: Every time I blink,
it changes from the house

1121
00:43:54,370 --> 00:43:54,870
to the face.

1122
00:43:54,870 --> 00:43:55,920
NANCY KANWISHER: Aha.

1123
00:43:55,920 --> 00:43:58,650
OK, it changes from the house
to the face only when you blink.

1124
00:43:58,650 --> 00:43:59,265
Yeah, Kwylie?

1125
00:43:59,265 --> 00:44:01,140
AUDIENCE: So at first
when I first put it on,

1126
00:44:01,140 --> 00:44:05,010
I was like I saw the house
and the face superimposed.

1127
00:44:05,010 --> 00:44:06,690
And then you told
us put one eye.

1128
00:44:06,690 --> 00:44:09,945
And then I opened both eyes,
and it went from the house

1129
00:44:09,945 --> 00:44:11,090
to the face.

1130
00:44:11,090 --> 00:44:12,090
NANCY KANWISHER: Uh huh.

1131
00:44:12,090 --> 00:44:12,660
Uh huh.

1132
00:44:12,660 --> 00:44:13,785
And then did it stay there?

1133
00:44:13,785 --> 00:44:16,080
Or does it keep flipping?

1134
00:44:16,080 --> 00:44:18,240
OK, if this is not working
for you, don't panic.

1135
00:44:18,240 --> 00:44:19,532
There's nothing wrong with you.

1136
00:44:19,532 --> 00:44:22,570
It's hard to get everybody's
red-green balance the same.

1137
00:44:22,570 --> 00:44:23,070
Yeah.

1138
00:44:23,070 --> 00:44:23,370
Sorry.

1139
00:44:23,370 --> 00:44:24,040
Question?

1140
00:44:24,040 --> 00:44:24,740
Shardul, yeah?

1141
00:44:24,740 --> 00:44:27,390
AUDIENCE: [INAUDIBLE]

1142
00:44:27,390 --> 00:44:29,550
NANCY KANWISHER: So I
share your intuition.

1143
00:44:29,550 --> 00:44:30,870
I think I can choose too.

1144
00:44:30,870 --> 00:44:33,040
But there's actually a
whole literature on that.

1145
00:44:33,040 --> 00:44:35,760
And the guy I was collaborating
with on this project,

1146
00:44:35,760 --> 00:44:38,220
20 years ago, insists
that, actually, that's

1147
00:44:38,220 --> 00:44:39,070
a wrong intuition.

1148
00:44:39,070 --> 00:44:39,780
You can't choose.

1149
00:44:39,780 --> 00:44:40,890
And I said, oh, the hell.

1150
00:44:40,890 --> 00:44:42,290
The hell I can't.

1151
00:44:42,290 --> 00:44:44,040
I'm in the scanner
being your best subject

1152
00:44:44,040 --> 00:44:45,810
ever because I'm
switching exactly when you

1153
00:44:45,810 --> 00:44:46,440
want me to switch.

1154
00:44:46,440 --> 00:44:47,400
Don't tell me I can't switch.

1155
00:44:47,400 --> 00:44:48,840
He said, no, the
literature says you don't.

1156
00:44:48,840 --> 00:44:50,760
Anyway, I don't know
the latest on that.

1157
00:44:50,760 --> 00:44:51,530
Yeah?

1158
00:44:51,530 --> 00:44:54,630
AUDIENCE: Kind of related,
it feels a lot harder

1159
00:44:54,630 --> 00:44:55,993
to see the house than the face.

1160
00:44:55,993 --> 00:44:56,910
NANCY KANWISHER: Yeah.

1161
00:44:56,910 --> 00:44:58,970
AUDIENCE: I can see the face
a lot easier than switching.

1162
00:44:58,970 --> 00:44:59,845
NANCY KANWISHER: Yes.

1163
00:44:59,845 --> 00:45:03,150
It's not optimally set up, so
that it's perfect for each.

1164
00:45:03,150 --> 00:45:05,880
Really the reason I do this
is to amuse myself looking

1165
00:45:05,880 --> 00:45:09,030
at all of you with your--

1166
00:45:09,030 --> 00:45:10,680
thank you.

1167
00:45:10,680 --> 00:45:12,262
OK, so you can keep
looking or not,

1168
00:45:12,262 --> 00:45:14,220
but I'm going to tell
you what's going on here.

1169
00:45:14,220 --> 00:45:16,290
The cool thing
about this is when--

1170
00:45:16,290 --> 00:45:18,230
if it didn't switch
for you, the person--

1171
00:45:18,230 --> 00:45:21,460
is there anybody for
whom it didn't switch?

1172
00:45:21,460 --> 00:45:21,960
One.

1173
00:45:21,960 --> 00:45:25,150
OK, so maybe your color
vision or who knows what.

1174
00:45:25,150 --> 00:45:26,640
But anyway.

1175
00:45:26,640 --> 00:45:29,610
OK, if it doesn't switch,
there's nothing wrong with you.

1176
00:45:29,610 --> 00:45:33,090
The percept everyone else has is
it just flips every few seconds

1177
00:45:33,090 --> 00:45:34,140
from one to the other.

1178
00:45:34,140 --> 00:45:34,980
OK.

1179
00:45:34,980 --> 00:45:38,130
So what's cool about this
is, when it switches,

1180
00:45:38,130 --> 00:45:40,350
nothing changes on your retina.

1181
00:45:40,350 --> 00:45:42,630
Only your state of
awareness changes.

1182
00:45:42,630 --> 00:45:45,240
And that gives us
a wonderful lever

1183
00:45:45,240 --> 00:45:48,750
to study what goes on in the
brain when awareness switches,

1184
00:45:48,750 --> 00:45:51,690
unconfounded from the stimulus.

1185
00:45:51,690 --> 00:45:52,680
OK?

1186
00:45:52,680 --> 00:45:55,470
It's like varying attention,
but it's more powerful.

1187
00:45:55,470 --> 00:45:58,800
So of course, we put
people in the scanner.

1188
00:45:58,800 --> 00:46:00,610
We put mostly me
and a few others.

1189
00:46:00,610 --> 00:46:01,170
But anyway.

1190
00:46:01,170 --> 00:46:03,030
We popped ourselves
in the scanner.

1191
00:46:03,030 --> 00:46:05,010
We just taped these
things on our forehead,

1192
00:46:05,010 --> 00:46:06,090
nice low-tech experiment.

1193
00:46:06,090 --> 00:46:07,980
And we looked at
that exact stimulus.

1194
00:46:07,980 --> 00:46:09,030
OK?

1195
00:46:09,030 --> 00:46:11,670
And we scanned the
brain while people

1196
00:46:11,670 --> 00:46:15,300
sat there watching the
stimulus flip back and forth.

1197
00:46:15,300 --> 00:46:15,960
OK?

1198
00:46:15,960 --> 00:46:18,630
So this is a stimulus
for the whole experiment.

1199
00:46:18,630 --> 00:46:19,740
This is the percept.

1200
00:46:19,740 --> 00:46:21,132
It switches back and forth.

1201
00:46:21,132 --> 00:46:23,340
The subject has a little
button box, so they can say,

1202
00:46:23,340 --> 00:46:24,330
now I see the face.

1203
00:46:24,330 --> 00:46:25,930
Now I see the house.

1204
00:46:25,930 --> 00:46:27,930
Now, of course, the choice
of a face and a house

1205
00:46:27,930 --> 00:46:29,520
was not random.

1206
00:46:29,520 --> 00:46:31,470
Why did we choose
a face and a house?

1207
00:46:31,470 --> 00:46:33,150
So that we could
look at the response

1208
00:46:33,150 --> 00:46:35,250
in the FFA and the PPA.

1209
00:46:35,250 --> 00:46:37,080
FFA loves faces, hates houses.

1210
00:46:37,080 --> 00:46:38,820
PPA loves houses, hates faces.

1211
00:46:38,820 --> 00:46:40,020
Perfect.

1212
00:46:40,020 --> 00:46:40,950
Right?

1213
00:46:40,950 --> 00:46:42,540
So the question
is, are they going

1214
00:46:42,540 --> 00:46:45,450
to switch when your percept
switches, even though nothing's

1215
00:46:45,450 --> 00:46:47,040
changing on your retina?

1216
00:46:47,040 --> 00:46:48,660
What do you guys think?

1217
00:46:48,660 --> 00:46:49,480
Switch?

1218
00:46:49,480 --> 00:46:49,980
Switch?

1219
00:46:49,980 --> 00:46:52,140
How many people think
it's going to switch?

1220
00:46:52,140 --> 00:46:54,430
How many people think
it's not going to switch?

1221
00:46:54,430 --> 00:46:55,300
A few.

1222
00:46:55,300 --> 00:46:58,420
OK, all right, so
that's what we did.

1223
00:46:58,420 --> 00:47:01,990
Here is now the raw MRI time
course averaged over the FFA

1224
00:47:01,990 --> 00:47:05,800
and averaged over the PPA for
one five-minute experiment

1225
00:47:05,800 --> 00:47:07,630
or three-minute
experiment in one subject.

1226
00:47:07,630 --> 00:47:08,170
Probably me.

1227
00:47:08,170 --> 00:47:09,520
I forget.

1228
00:47:09,520 --> 00:47:11,380
Again, stimulus is the
same the whole time.

1229
00:47:11,380 --> 00:47:15,340
The letters are the times when
the subject pressed the button.

1230
00:47:15,340 --> 00:47:17,320
And you can sort of get
a sense that there's

1231
00:47:17,320 --> 00:47:20,770
a little bit of maybe like
here's a phase the FFA

1232
00:47:20,770 --> 00:47:21,940
response goes up.

1233
00:47:21,940 --> 00:47:24,430
Here again the house goes up.

1234
00:47:24,430 --> 00:47:26,140
But it's kind of hard to tell.

1235
00:47:26,140 --> 00:47:28,420
So really, the way
to analyze these data

1236
00:47:28,420 --> 00:47:33,250
is to take all the face
to house flips, right,

1237
00:47:33,250 --> 00:47:36,770
and align them and signal
average to get rid of noise.

1238
00:47:36,770 --> 00:47:37,270
OK?

1239
00:47:37,270 --> 00:47:41,710
So we can look at, what happens
before and after a switch

1240
00:47:41,710 --> 00:47:42,910
from face to house?

1241
00:47:42,910 --> 00:47:45,100
And what happens in the
opposite direction switch?

1242
00:47:45,100 --> 00:47:46,580
Everybody get the idea?

1243
00:47:46,580 --> 00:47:49,090
It's just a way to
clean up the noise here.

1244
00:47:49,090 --> 00:47:50,410
And here's what happens.

1245
00:47:50,410 --> 00:47:51,500
This is time zero.

1246
00:47:51,500 --> 00:47:53,680
The subject presses
a button saying

1247
00:47:53,680 --> 00:47:57,280
that their percept has flipped
between a house to a face.

1248
00:47:57,280 --> 00:47:59,620
The response in
the FFA shoots up.

1249
00:47:59,620 --> 00:48:03,355
And the response in
the PPA shoots down.

1250
00:48:03,355 --> 00:48:05,430
Isn't that cool?

1251
00:48:05,430 --> 00:48:10,860
OK, now why are these peaks and
valleys after the button press?

1252
00:48:14,280 --> 00:48:14,780
Yeah.

1253
00:48:14,780 --> 00:48:15,320
Evan, right.

1254
00:48:15,320 --> 00:48:17,450
There's a delayed signal, right?

1255
00:48:17,450 --> 00:48:19,640
OK?

1256
00:48:19,640 --> 00:48:20,630
This is harder.

1257
00:48:20,630 --> 00:48:22,880
Why do they come back
together out there?

1258
00:48:27,060 --> 00:48:28,830
Yeah, because it flips back.

1259
00:48:28,830 --> 00:48:30,300
It flips back.

1260
00:48:30,300 --> 00:48:33,270
And all we've done is signal
average to one button press.

1261
00:48:33,270 --> 00:48:35,320
But there are different
durations of percepts.

1262
00:48:35,320 --> 00:48:36,900
Sometimes it lasts
three seconds.

1263
00:48:36,900 --> 00:48:39,480
Sometimes it lasts 10 seconds,
everything in between.

1264
00:48:39,480 --> 00:48:41,250
By the time you're
12 seconds out,

1265
00:48:41,250 --> 00:48:42,750
most people have flipped again.

1266
00:48:42,750 --> 00:48:44,550
That's why it goes back.

1267
00:48:44,550 --> 00:48:45,435
OK, make sense?

1268
00:48:48,420 --> 00:48:49,237
Yeah?

1269
00:48:49,237 --> 00:49:05,040
AUDIENCE: [INAUDIBLE]

1270
00:49:05,040 --> 00:49:06,290
NANCY KANWISHER: I don't know.

1271
00:49:06,290 --> 00:49:09,050
And I'm not sure that
that was generally true.

1272
00:49:09,050 --> 00:49:10,670
I hadn't noticed that before.

1273
00:49:10,670 --> 00:49:12,830
But we should
scrutinize the images.

1274
00:49:12,830 --> 00:49:16,040
I can't actually remember if
this was averaged over subjects

1275
00:49:16,040 --> 00:49:17,150
or if this is one subject.

1276
00:49:17,150 --> 00:49:18,530
I suspect that's a fluke.

1277
00:49:18,530 --> 00:49:21,640
But we could look at it and see.

1278
00:49:21,640 --> 00:49:26,310
OK, so the point of all of this
is that these regions, the face

1279
00:49:26,310 --> 00:49:28,590
area and the place area,
care about what you

1280
00:49:28,590 --> 00:49:31,290
are experiencing,
unconfounded from what's

1281
00:49:31,290 --> 00:49:33,120
hitting your retina, right?

1282
00:49:33,120 --> 00:49:35,620
They are reflecting the
contents of your awareness,

1283
00:49:35,620 --> 00:49:41,370
not just what's coming into
your eyes, which is cool.

1284
00:49:41,370 --> 00:49:45,430
OK, so that shows that
these regions kind

1285
00:49:45,430 --> 00:49:47,710
of track awareness.

1286
00:49:47,710 --> 00:49:48,670
OK?

1287
00:49:48,670 --> 00:49:52,330
But can we find any
evidence for perception

1288
00:49:52,330 --> 00:49:55,630
without awareness
using neuroimaging?

1289
00:49:55,630 --> 00:49:56,560
OK?

1290
00:49:56,560 --> 00:49:59,230
So to make this point, I'm
going to accelerate slightly,

1291
00:49:59,230 --> 00:50:00,968
but I think there's
just barely time.

1292
00:50:00,968 --> 00:50:03,010
There are lots and lots
of studies of this genre.

1293
00:50:03,010 --> 00:50:04,750
And I'll tell you about one.

1294
00:50:04,750 --> 00:50:05,800
So I'm going to show you.

1295
00:50:05,800 --> 00:50:08,110
I'm going to flash up a
very rapid series of digits.

1296
00:50:08,110 --> 00:50:10,600
You have to get ready
to write stuff down.

1297
00:50:10,600 --> 00:50:15,610
Your task is to see if there
are any letters in the digit.

1298
00:50:15,610 --> 00:50:16,480
There might be zero.

1299
00:50:16,480 --> 00:50:17,200
There might be one.

1300
00:50:17,200 --> 00:50:17,920
There might be a few.

1301
00:50:17,920 --> 00:50:19,330
Write down any letters you see.

1302
00:50:19,330 --> 00:50:20,960
It's going to go by really fast.

1303
00:50:20,960 --> 00:50:23,500
OK, everyone ready?

1304
00:50:23,500 --> 00:50:25,930
OK.

1305
00:50:25,930 --> 00:50:26,500
Here we go.

1306
00:50:30,080 --> 00:50:33,320
OK, write down any
letters you may have seen.

1307
00:50:33,320 --> 00:50:35,180
OK?

1308
00:50:35,180 --> 00:50:37,640
All right, let's do another one.

1309
00:50:37,640 --> 00:50:38,390
Everyone ready?

1310
00:50:43,630 --> 00:50:47,140
OK, write down any
letters you may have seen.

1311
00:50:47,140 --> 00:50:48,480
OK?

1312
00:50:48,480 --> 00:50:52,560
OK, raise your hand if you
saw both the A and the P

1313
00:50:52,560 --> 00:50:55,030
in the first sequence.

1314
00:50:55,030 --> 00:50:55,630
A few of you.

1315
00:50:55,630 --> 00:50:57,130
Less than half.

1316
00:50:57,130 --> 00:51:00,100
Raise your hand if you saw both
the X and the H in the second.

1317
00:51:00,100 --> 00:51:01,600
Almost everyone.

1318
00:51:01,600 --> 00:51:03,670
All right, well, I
cheated slightly.

1319
00:51:03,670 --> 00:51:05,650
But never mind how I cheated.

1320
00:51:05,650 --> 00:51:07,150
The way you cheat in demos--

1321
00:51:07,150 --> 00:51:08,950
you guys will be
teaching someday.

1322
00:51:08,950 --> 00:51:11,630
You have the one you want
people to get wrong first.

1323
00:51:11,630 --> 00:51:12,610
It's a total cheat.

1324
00:51:12,610 --> 00:51:13,660
So it's slightly cheated.

1325
00:51:13,660 --> 00:51:15,718
But if you do it right--

1326
00:51:15,718 --> 00:51:16,760
it didn't work last year.

1327
00:51:16,760 --> 00:51:17,635
That's why I cheated.

1328
00:51:17,635 --> 00:51:19,510
Anyway, even if you do
it right in the lab,

1329
00:51:19,510 --> 00:51:22,540
you get a really strong effect.

1330
00:51:22,540 --> 00:51:24,910
There was only one
digit between the A

1331
00:51:24,910 --> 00:51:26,950
and the P in the first sequence.

1332
00:51:26,950 --> 00:51:29,620
And it's like your brain
is still dealing with the A

1333
00:51:29,620 --> 00:51:30,983
when the P comes along.

1334
00:51:30,983 --> 00:51:33,400
You just don't even see the
P. You're looking for letters.

1335
00:51:33,400 --> 00:51:34,750
You don't see it.

1336
00:51:34,750 --> 00:51:36,910
Whereas there were three
digits between the X

1337
00:51:36,910 --> 00:51:39,370
and the H in the
second sequence.

1338
00:51:39,370 --> 00:51:41,930
And just to show
you some real data.

1339
00:51:41,930 --> 00:51:44,030
So this has been done
lots of different ways.

1340
00:51:44,030 --> 00:51:46,090
Here's an example
of this experiment.

1341
00:51:46,090 --> 00:51:50,738
And this is the time that people
get the second letter when they

1342
00:51:50,738 --> 00:51:52,030
don't have to report the first.

1343
00:51:52,030 --> 00:51:53,350
Well, the first
one's colored, right?

1344
00:51:53,350 --> 00:51:55,480
So you either have to
report the colored letter

1345
00:51:55,480 --> 00:51:57,430
and detect the letter
after it or just

1346
00:51:57,430 --> 00:51:59,990
detect the letter after
the colored thing.

1347
00:51:59,990 --> 00:52:02,120
So if you don't have to
report the first one,

1348
00:52:02,120 --> 00:52:03,530
there's no dip in accuracy.

1349
00:52:03,530 --> 00:52:04,030
Oh, sorry.

1350
00:52:04,030 --> 00:52:06,340
This is a function of the
distance between the two

1351
00:52:06,340 --> 00:52:08,000
items in the sequence.

1352
00:52:08,000 --> 00:52:10,570
But if you do,
there's a big dip.

1353
00:52:10,570 --> 00:52:13,810
That's maximal with
one intervening item.

1354
00:52:13,810 --> 00:52:15,470
Yeah?

1355
00:52:15,470 --> 00:52:19,970
OK, well, here it
says two or three.

1356
00:52:19,970 --> 00:52:22,370
But anyway.

1357
00:52:22,370 --> 00:52:24,525
OK, so that's called
the attentional blink.

1358
00:52:24,525 --> 00:52:26,900
And the idea is there isn't
a physical blink of the eyes.

1359
00:52:26,900 --> 00:52:28,793
But your attention
system is tied up

1360
00:52:28,793 --> 00:52:29,960
processing the first target.

1361
00:52:29,960 --> 00:52:32,300
And it doesn't get
the second one, OK?

1362
00:52:32,300 --> 00:52:34,332
So there are dozens,
probably hundreds

1363
00:52:34,332 --> 00:52:35,540
of papers on this phenomenon.

1364
00:52:35,540 --> 00:52:37,190
It's pretty cool.

1365
00:52:37,190 --> 00:52:41,450
But for present purposes,
we're going to use it to say,

1366
00:52:41,450 --> 00:52:45,170
does that unseen second
target get into the brain?

1367
00:52:45,170 --> 00:52:46,610
Presumably, it
lands on the retina

1368
00:52:46,610 --> 00:52:47,808
because you didn't blink.

1369
00:52:47,808 --> 00:52:49,100
So it could probably get there.

1370
00:52:49,100 --> 00:52:50,210
Probably got to V1.

1371
00:52:50,210 --> 00:52:53,930
How far up the
system did it get?

1372
00:52:53,930 --> 00:52:58,130
Can you think of how we might
test this using functional MRI.

1373
00:52:58,130 --> 00:53:02,090
How would we see how
far those stimuli

1374
00:53:02,090 --> 00:53:03,890
go in an experiment like this?

1375
00:53:07,750 --> 00:53:08,500
What would we use?

1376
00:53:08,500 --> 00:53:11,770
What would we measure
the MRI response to?

1377
00:53:11,770 --> 00:53:13,427
How could we design
a version of this

1378
00:53:13,427 --> 00:53:15,010
that you could do
with functional MRI?

1379
00:53:19,690 --> 00:53:24,040
We need some MRI response,
where, if we get that response,

1380
00:53:24,040 --> 00:53:26,455
we know that it's a response
to a given stimulus.

1381
00:53:29,560 --> 00:53:31,990
What would we measure?

1382
00:53:31,990 --> 00:53:33,250
And what stimuli would we use?

1383
00:53:35,830 --> 00:53:37,953
AUDIENCE: Well, we could
use faces or something.

1384
00:53:37,953 --> 00:53:38,870
NANCY KANWISHER: Yeah.

1385
00:53:38,870 --> 00:53:39,920
It's always the same.

1386
00:53:39,920 --> 00:53:40,460
Yeah.

1387
00:53:40,460 --> 00:53:41,930
Faces and houses.

1388
00:53:41,930 --> 00:53:43,190
Lots of ways to do this.

1389
00:53:43,190 --> 00:53:46,987
But it's not even my experiment,
and they used faces and houses.

1390
00:53:46,987 --> 00:53:49,070
OK, so they did a version
of that very same thing.

1391
00:53:49,070 --> 00:53:51,230
This is a bunch of
garbage flashing on.

1392
00:53:51,230 --> 00:53:52,470
Early on, there's a face.

1393
00:53:52,470 --> 00:53:54,345
The subjects have to
say which of those three

1394
00:53:54,345 --> 00:53:55,920
faces they've studied it is.

1395
00:53:55,920 --> 00:53:58,520
And then after that,
a scene comes on,

1396
00:53:58,520 --> 00:54:00,290
or it doesn't come on.

1397
00:54:00,290 --> 00:54:03,110
OK, they have to just
say, was there a scene?

1398
00:54:03,110 --> 00:54:04,790
First of all, which face was it?

1399
00:54:04,790 --> 00:54:07,350
And was there a
scene afterwards?

1400
00:54:07,350 --> 00:54:10,000
OK, so it's a variant of
the thing you just did.

1401
00:54:10,000 --> 00:54:12,810
So here's the behavioral data.

1402
00:54:12,810 --> 00:54:14,760
If you don't have
to report the face,

1403
00:54:14,760 --> 00:54:17,580
then you are very accurate
reporting the scene.

1404
00:54:17,580 --> 00:54:19,380
If you do have to
report the face,

1405
00:54:19,380 --> 00:54:22,980
then you're very bad at it if
the scene comes right after.

1406
00:54:22,980 --> 00:54:24,690
But if there's a big
interval in between,

1407
00:54:24,690 --> 00:54:25,740
you do OK with the scene.

1408
00:54:25,740 --> 00:54:28,260
Same thing we saw before but
now with faces and scenes.

1409
00:54:28,260 --> 00:54:30,390
OK, that's the behavioral data.

1410
00:54:30,390 --> 00:54:32,700
What do you see
in the PPA, right?

1411
00:54:32,700 --> 00:54:34,920
The second target
is always a scene.

1412
00:54:34,920 --> 00:54:37,500
So what you see
in the PPA, here's

1413
00:54:37,500 --> 00:54:42,090
a response in the PPA to that
second scene if it's a hit.

1414
00:54:42,090 --> 00:54:43,680
That is if you detect it, right?

1415
00:54:43,680 --> 00:54:46,500
You correctly say, yes, there
was a scene in that trial.

1416
00:54:46,500 --> 00:54:48,490
Here's a response
with a correct reject.

1417
00:54:48,490 --> 00:54:49,950
There was no scene
in that trial.

1418
00:54:49,950 --> 00:54:52,290
And you correctly
said there was not.

1419
00:54:52,290 --> 00:54:55,110
OK, so that difference
shows you the response

1420
00:54:55,110 --> 00:54:58,320
in the PPA to a scene
that you've seen,

1421
00:54:58,320 --> 00:55:00,360
that you detected consciously.

1422
00:55:00,360 --> 00:55:02,902
But here's a critical case.

1423
00:55:02,902 --> 00:55:04,860
I don't know why there's
a little pop up there.

1424
00:55:04,860 --> 00:55:08,250
Anyway, this is a case
where there was a scene,

1425
00:55:08,250 --> 00:55:10,370
but you've said there wasn't.

1426
00:55:10,370 --> 00:55:10,870
OK?

1427
00:55:10,870 --> 00:55:11,745
That's called a miss.

1428
00:55:11,745 --> 00:55:12,370
You missed it.

1429
00:55:12,370 --> 00:55:14,710
It was there, and
you didn't see it.

1430
00:55:14,710 --> 00:55:17,830
And what it shows you
is two cool things.

1431
00:55:17,830 --> 00:55:21,490
Well, first of all, it tells
you that the more aware you

1432
00:55:21,490 --> 00:55:24,010
are of the stimulus,
the stronger response.

1433
00:55:24,010 --> 00:55:27,120
But the crucial
thing it shows you

1434
00:55:27,120 --> 00:55:29,880
is this difference right
here, which is significant.

1435
00:55:29,880 --> 00:55:33,030
This is a kind of perception
without awareness.

1436
00:55:33,030 --> 00:55:34,650
You did not detect the scene.

1437
00:55:34,650 --> 00:55:36,000
You said there was no scene.

1438
00:55:36,000 --> 00:55:38,850
But your PPA detected
the difference.

1439
00:55:38,850 --> 00:55:40,930
Everybody get that?

1440
00:55:40,930 --> 00:55:42,900
So we have evidence
for perception

1441
00:55:42,900 --> 00:55:47,160
by the PPA at least of a
scene without awareness.

1442
00:55:47,160 --> 00:55:48,730
That make sense?

1443
00:55:48,730 --> 00:55:51,650
So there are many, many
studies like this that kind of

1444
00:55:51,650 --> 00:55:52,650
take it every which way.

1445
00:55:52,650 --> 00:55:54,630
This is just one
little example of how

1446
00:55:54,630 --> 00:55:58,200
you can use neuroimaging to
ask, how far up the system

1447
00:55:58,200 --> 00:56:01,110
does an unseen stimulus get?

1448
00:56:01,110 --> 00:56:03,430
Make sense?

1449
00:56:03,430 --> 00:56:05,800
OK, so we are about out of time.

1450
00:56:05,800 --> 00:56:07,870
And I just gave
you a little taste

1451
00:56:07,870 --> 00:56:10,390
of some of the work on neural
correlates of awareness,

1452
00:56:10,390 --> 00:56:13,960
showed you, with binocular
rivalry, a neural response

1453
00:56:13,960 --> 00:56:17,410
that's correlated with awareness
unconfounded from the stimulus,

1454
00:56:17,410 --> 00:56:20,020
uncoupling, again, perceptual
awareness from what's

1455
00:56:20,020 --> 00:56:23,650
hitting your retina, and, in the
case of the attentional blink,

1456
00:56:23,650 --> 00:56:27,250
some evidence for
perceptual representation

1457
00:56:27,250 --> 00:56:29,150
without awareness.

1458
00:56:29,150 --> 00:56:30,700
OK?