1
00:00:00,000 --> 00:00:02,510
ANNOUNCER: The following content
is provided by MIT

2
00:00:02,510 --> 00:00:05,231
OpenCourseWare under a Creative
Commons license.

3
00:00:05,231 --> 00:00:08,747
Additional information about
our license, and MIT

4
00:00:08,747 --> 00:00:12,263
OpenCourseWare in general, is
available at ocw.mit.edu.

5
00:00:15,780 --> 00:00:16,240
PROFESSOR: Good afternoon.

6
00:00:16,240 --> 00:00:17,490
AUDIENCE: Good afternoon.

7
00:00:25,030 --> 00:00:30,100
PROFESSOR: So there I was in
my car this morning as the

8
00:00:30,100 --> 00:00:36,230
pouring rain started, thinking
if I make a dash for it --

9
00:00:36,230 --> 00:00:37,580
I've got to take the computer.

10
00:00:37,580 --> 00:00:39,730
I need the coffee, because
otherwise I'm

11
00:00:39,730 --> 00:00:41,660
going to fall asleep.

12
00:00:41,660 --> 00:00:45,190
I don't need anything
in that bag, do I?

13
00:00:45,190 --> 00:00:46,980
So I took off without the bag.

14
00:00:46,980 --> 00:00:49,680
And it was true that I
didn't need most of

15
00:00:49,680 --> 00:00:50,560
what was in the bag.

16
00:00:50,560 --> 00:00:52,860
But the lecture notes for
today's lecture would have

17
00:00:52,860 --> 00:00:55,580
been a useful thing
to take with me.

18
00:00:55,580 --> 00:00:58,880
On the other hand, if there
was ever going to be a day

19
00:00:58,880 --> 00:01:01,560
where I forgot the lecture
notes, this is probably the

20
00:01:01,560 --> 00:01:02,310
one to do it.

21
00:01:02,310 --> 00:01:06,240
Because I'm going to talk
about attention today.

22
00:01:06,240 --> 00:01:11,210
And attention research is
what I do for a living.

23
00:01:11,210 --> 00:01:14,780
If there's anything that I
should be able to just stand

24
00:01:14,780 --> 00:01:18,220
up and lecture about,
this is it.

25
00:01:18,220 --> 00:01:22,070
Now of course, that means that
you should find this to be the

26
00:01:22,070 --> 00:01:26,190
most gripping topic in the
entire course, and that you

27
00:01:26,190 --> 00:01:30,990
should decide that you want
to do this for a living.

28
00:01:30,990 --> 00:01:33,260
Well, when you decide you want
to do it for a sort of a

29
00:01:33,260 --> 00:01:40,110
living, like a $10 an hour
living, you can come and be a

30
00:01:40,110 --> 00:01:43,050
subject in attention
research in my lab.

31
00:01:43,050 --> 00:01:45,620
I would again advocate
that you sign up --

32
00:01:45,620 --> 00:01:50,020
I saw there are still these
notes around about signing up

33
00:01:50,020 --> 00:01:51,940
to be a subject generally
in BCS.

34
00:01:51,940 --> 00:01:54,720
My lab is separate from the
BCS business because I'm

35
00:01:54,720 --> 00:01:57,240
technically Brigham and
Women's Hospital.

36
00:01:57,240 --> 00:02:00,430
But you can sign up with us,
too, and we'll pay you $10 an

37
00:02:00,430 --> 00:02:03,730
hour to do visual attention
research.

38
00:02:03,730 --> 00:02:05,760
What could be better
than that?

39
00:02:05,760 --> 00:02:07,810
Is Kristen here?

40
00:02:07,810 --> 00:02:10,400
I don't see Kristen.

41
00:02:10,400 --> 00:02:14,490
Kristen, one of the TAs is
also in my lab, and I was

42
00:02:14,490 --> 00:02:16,030
going to point her out.

43
00:02:16,030 --> 00:02:20,050
Anyway, send me an email,
we'll sign you up.

44
00:02:20,050 --> 00:02:21,770
Talk to me.

45
00:02:21,770 --> 00:02:23,440
We'd love to have you.

46
00:02:23,440 --> 00:02:27,450
And you can do Where's Waldo
experiments for $10 an hour.

47
00:02:29,950 --> 00:02:31,650
You think I'm joking.

48
00:02:38,650 --> 00:02:43,530
Let me try to explain why it
is that I'm putting an

49
00:02:43,530 --> 00:02:51,040
attention lecture in between
a sensation lecture and a

50
00:02:51,040 --> 00:02:52,150
perception lecture.

51
00:02:52,150 --> 00:02:53,810
It's not terribly typical.

52
00:02:53,810 --> 00:02:57,070
More typically, if people talk
about attention, they go off

53
00:02:57,070 --> 00:03:02,320
and do it later, after doing
sensation and perception.

54
00:03:02,320 --> 00:03:05,450
But why am I putting
it in there?

55
00:03:05,450 --> 00:03:12,630
The core reason is that you
simply cannot process all of

56
00:03:12,630 --> 00:03:15,250
the information that you
take in from the world.

57
00:03:15,250 --> 00:03:19,000
You're taking in a vast amount
of sensory information.

58
00:03:19,000 --> 00:03:23,790
Your perceptual capabilities
-- for instance, those that

59
00:03:23,790 --> 00:03:26,450
allow you to recognize
specific objects --

60
00:03:26,450 --> 00:03:27,440
are limited.

61
00:03:27,440 --> 00:03:30,910
You cannot recognize all of the
objects in the world that

62
00:03:30,910 --> 00:03:32,570
you are looking at, all
at the same time.

63
00:03:32,570 --> 00:03:33,860
It simply doesn't work.

64
00:03:33,860 --> 00:03:43,180
And so, roughly speaking,
there's the situation.

65
00:03:43,180 --> 00:03:49,340
You've got a lot of stuff coming
in from the outside.

66
00:03:49,340 --> 00:03:55,230
And you've a box here that does,
let's say, let's call

67
00:03:55,230 --> 00:03:59,590
this one a recognition box.

68
00:03:59,590 --> 00:04:06,060
And only one thing at a time
gets to go in and come out of

69
00:04:06,060 --> 00:04:08,030
that box, basically.

70
00:04:08,030 --> 00:04:13,820
So this is like the basic MIT
metaphor about drinking from

71
00:04:13,820 --> 00:04:16,480
the firehose.

72
00:04:16,480 --> 00:04:20,090
Well, if you're really going
to drink from the firehose,

73
00:04:20,090 --> 00:04:23,560
it's a very useful idea to
restrict the flow in some

74
00:04:23,560 --> 00:04:26,280
fashion, and let some of
that water just [SPLAT]

75
00:04:26,280 --> 00:04:28,210
and get you wet, or whatever.

76
00:04:28,210 --> 00:04:32,920
And so there's a severe
constriction, sometimes called

77
00:04:32,920 --> 00:04:33,690
a bottleneck --

78
00:04:33,690 --> 00:04:35,270
I think I've got some slides
that call it a

79
00:04:35,270 --> 00:04:37,500
bottleneck later --

80
00:04:37,500 --> 00:04:40,650
that takes all of this and only
lets some of it through.

81
00:04:40,650 --> 00:04:44,250
And that bottleneck is governed
-- it's not just

82
00:04:44,250 --> 00:04:47,400
random what gets through -- it's
governed by mechanisms of

83
00:04:47,400 --> 00:04:52,280
selective attention that allow
some things to get through and

84
00:04:52,280 --> 00:04:57,660
leave other things on
the on the floor.

85
00:04:57,660 --> 00:05:00,340
And so if you think of this
as sort of sensation and

86
00:05:00,340 --> 00:05:02,950
perception -- which is
a little bald, but --

87
00:05:02,950 --> 00:05:05,490
then that's why you put
attention in the middle there.

88
00:05:05,490 --> 00:05:08,860
Now to motivate this a bit
further, let me do a

89
00:05:08,860 --> 00:05:09,870
demonstration.

90
00:05:09,870 --> 00:05:13,250
Actually, this is the
demonstration of why reading

91
00:05:13,250 --> 00:05:16,480
the Tech while listening
to my lecture may not

92
00:05:16,480 --> 00:05:19,100
be a brilliant idea.

93
00:05:19,100 --> 00:05:20,660
Well, it may be a
brilliant idea.

94
00:05:20,660 --> 00:05:24,060
It just depends on your
particular goals in life.

95
00:05:24,060 --> 00:05:27,910
I need a couple of volunteer
type people who

96
00:05:27,910 --> 00:05:31,200
wish to read here.

97
00:05:31,200 --> 00:05:35,250
All right, there's a volunteer
person, and there's a pink

98
00:05:35,250 --> 00:05:36,150
volunteer person.

99
00:05:36,150 --> 00:05:38,700
Yes, you, MIT person.

100
00:05:38,700 --> 00:05:39,810
But you have to come up here.

101
00:05:39,810 --> 00:05:44,620
So kick a few people on the way
by and stuff like that.

102
00:05:44,620 --> 00:05:46,820
You have to come up here and
do a dramatic reading.

103
00:05:52,520 --> 00:06:02,070
What I'm going to do is have
these people both read to you

104
00:06:02,070 --> 00:06:05,070
at the same time.

105
00:06:05,070 --> 00:06:10,010
You're going to read from
here, from where it says

106
00:06:10,010 --> 00:06:18,160
"Catherine." And you're going
to read from here.

107
00:06:18,160 --> 00:06:23,850
And you're both going to read
nice and loudly and steadily.

108
00:06:23,850 --> 00:06:25,420
At the same time, yes.

109
00:06:25,420 --> 00:06:27,880
That's the interesting part.

110
00:06:27,880 --> 00:06:31,370
And what you're going to do is
you're going to listen to her

111
00:06:31,370 --> 00:06:35,360
-- for, actually, to
"her" specifically.

112
00:06:35,360 --> 00:06:40,090
Listen for the third instance
of the word "her." When you

113
00:06:40,090 --> 00:06:45,180
hear "her," say "her." For the
third time, raise your hand.

114
00:06:45,180 --> 00:06:47,190
OK?

115
00:06:47,190 --> 00:06:48,860
Yeah, we got this?

116
00:06:48,860 --> 00:06:50,550
Yeah, all right.

117
00:06:50,550 --> 00:06:51,880
This her, not that her.

118
00:06:51,880 --> 00:06:53,580
NINA: I could tell
you my name.

119
00:06:53,580 --> 00:06:54,170
PROFESSOR: That would help.

120
00:06:54,170 --> 00:06:54,825
You are?

121
00:06:54,825 --> 00:06:55,340
NINA: Nina.

122
00:06:55,340 --> 00:06:55,880
PROFESSOR: That's Nina.

123
00:06:55,880 --> 00:06:56,410
This is?

124
00:06:56,410 --> 00:06:57,426
ZAINA: Zaina.

125
00:06:57,426 --> 00:06:59,530
[LAUGHTER]

126
00:06:59,530 --> 00:07:00,100
PROFESSOR: Right.

127
00:07:00,100 --> 00:07:02,000
OK.

128
00:07:02,000 --> 00:07:03,360
Zaina or Zena?

129
00:07:03,360 --> 00:07:03,630
ZAINA: Zaina.

130
00:07:03,630 --> 00:07:04,000
PROFESSOR: OK.

131
00:07:04,000 --> 00:07:05,570
At least it's not
just one letter.

132
00:07:05,570 --> 00:07:06,960
NINA: My surname's
[? Navarre, ?] if that helps.

133
00:07:06,960 --> 00:07:09,320
PROFESSOR: No, this is not
going to help at all.

134
00:07:09,320 --> 00:07:11,390
Her.

135
00:07:11,390 --> 00:07:14,070
When Nina says "her" for the
third time, raise your hand.

136
00:07:14,070 --> 00:07:15,900
OK?

137
00:07:15,900 --> 00:07:17,210
You got it?

138
00:07:17,210 --> 00:07:18,010
You got it?

139
00:07:18,010 --> 00:07:22,260
On your mark, get set, read.

140
00:07:22,260 --> 00:07:33,230
[OVERLAPPING VOICES]

141
00:07:33,230 --> 00:07:35,540
PROFESSOR: OK, thank you.

142
00:07:35,540 --> 00:07:39,920
All right, that was excellent.

143
00:07:39,920 --> 00:07:42,650
All right, so what was
she talking about?

144
00:07:42,650 --> 00:07:43,850
Yeah, something.

145
00:07:43,850 --> 00:07:46,570
No, no, somebody raise
their hand.

146
00:07:46,570 --> 00:07:47,510
Hand, hand.

147
00:07:47,510 --> 00:07:48,470
What was --?

148
00:07:48,470 --> 00:07:49,350
A letter, thank you.

149
00:07:49,350 --> 00:07:51,270
That sounds good.

150
00:07:51,270 --> 00:07:52,350
She'd gotten a letter.

151
00:07:52,350 --> 00:07:54,350
Was it a nice letter?

152
00:07:54,350 --> 00:07:54,730
Who knows?

153
00:07:54,730 --> 00:07:55,660
It didn't sound too good.

154
00:07:55,660 --> 00:07:57,330
Her countenance wasn't
doing good things.

155
00:07:57,330 --> 00:07:59,370
What was she talking about?

156
00:07:59,370 --> 00:08:00,360
Zaina?

157
00:08:00,360 --> 00:08:01,610
What?

158
00:08:03,300 --> 00:08:04,240
Oh, uh.

159
00:08:04,240 --> 00:08:05,110
She was talking about uh.

160
00:08:05,110 --> 00:08:06,770
OK.

161
00:08:06,770 --> 00:08:08,000
Was she talking?

162
00:08:08,000 --> 00:08:09,070
AUDIENCE: Yes.

163
00:08:09,070 --> 00:08:12,010
PROFESSOR: So what's
your problem?

164
00:08:12,010 --> 00:08:14,400
What you were doing was --

165
00:08:14,400 --> 00:08:17,245
so how many people -- well,
obviously the hands suggested

166
00:08:17,245 --> 00:08:20,050
that everybody could manage
to do the task.

167
00:08:20,050 --> 00:08:22,260
What could you pick up from --

168
00:08:22,260 --> 00:08:22,430
Zena?

169
00:08:22,430 --> 00:08:23,440
ZAINA: Zaina.

170
00:08:23,440 --> 00:08:24,760
PROFESSOR: Zaina.

171
00:08:24,760 --> 00:08:26,180
I'm not going to -- otherwise
it's going to turn into

172
00:08:26,180 --> 00:08:29,310
warrior queen and
stuff like that.

173
00:08:29,310 --> 00:08:38,070
What could you pick up
about Zaina's speech?

174
00:08:38,070 --> 00:08:39,860
Anything?

175
00:08:39,860 --> 00:08:40,670
No content.

176
00:08:40,670 --> 00:08:43,240
How many people knew
she was talking?

177
00:08:43,240 --> 00:08:44,440
All right, so you can
pick up something.

178
00:08:44,440 --> 00:08:45,530
What else did you
know about her?

179
00:08:45,530 --> 00:08:47,830
AUDIENCE: The tone she
was speaking in.

180
00:08:47,830 --> 00:08:49,395
PROFESSOR: The tone she
was speaking in.

181
00:08:49,395 --> 00:08:52,230
If it'd been a male voice, if
she had switched to a male

182
00:08:52,230 --> 00:08:54,520
voice, you would've noticed.

183
00:08:54,520 --> 00:08:55,620
Anything else, you think?

184
00:08:55,620 --> 00:08:57,850
AUDIENCE: Was she reading
from Heart of Darkness?

185
00:08:57,850 --> 00:09:01,630
PROFESSOR: Was reading from
Heart of Darkness?

186
00:09:01,630 --> 00:09:05,660
No, actually what she was
reading from was Lucretius, On

187
00:09:05,660 --> 00:09:06,980
the Nature Of the Universe.

188
00:09:06,980 --> 00:09:08,240
A wonderful book.

189
00:09:08,240 --> 00:09:10,280
De Rerum Natura in Latin.

190
00:09:10,280 --> 00:09:11,690
He's a Roman author.

191
00:09:11,690 --> 00:09:14,620
This is sort of the first
intro psych book.

192
00:09:14,620 --> 00:09:17,060
It's also the first intro
physics book, intro

193
00:09:17,060 --> 00:09:18,250
everything.

194
00:09:18,250 --> 00:09:20,330
In those days, you could write a
book called On the Nature Of

195
00:09:20,330 --> 00:09:23,460
the Universe, in verse.

196
00:09:23,460 --> 00:09:26,060
This is a prose translation.

197
00:09:26,060 --> 00:09:30,300
She was actually reading
Lucretius' theory of vision.

198
00:09:30,300 --> 00:09:32,730
And even she may not have
noticed that, because it's all

199
00:09:32,730 --> 00:09:35,410
about thin films and cool
stuff like that.

200
00:09:35,410 --> 00:09:36,760
AUDIENCE: [INAUDIBLE] video.

201
00:09:36,760 --> 00:09:37,550
PROFESSOR: A video.

202
00:09:37,550 --> 00:09:41,170
Yeah, well it's an ancient
Roman video.

203
00:09:41,170 --> 00:09:45,140
But only a very limited amount
of stuff got in.

204
00:09:45,140 --> 00:09:48,210
So there was a certain amount of
stuff that was getting in.

205
00:09:48,210 --> 00:09:52,160
But at some point your auditory
system gave up on

206
00:09:52,160 --> 00:09:54,400
processing that stream.

207
00:09:54,400 --> 00:10:00,180
And in terms of extracting
meaning, understanding the

208
00:10:00,180 --> 00:10:04,120
words, it went with Nina,
because that was the job.

209
00:10:04,120 --> 00:10:06,580
You can't do both of them.

210
00:10:06,580 --> 00:10:07,710
We'd better let them
go sit down.

211
00:10:07,710 --> 00:10:09,220
Thank you for being --

212
00:10:13,050 --> 00:10:15,350
What would have made
the task easier?

213
00:10:15,350 --> 00:10:18,070
What would make it easier to pay
attention to one and not

214
00:10:18,070 --> 00:10:18,970
the other, do you think?

215
00:10:18,970 --> 00:10:20,330
AUDIENCE: Amplifying
one of them.

216
00:10:20,330 --> 00:10:21,330
PROFESSOR: Amplifying
one of them.

217
00:10:21,330 --> 00:10:25,160
Yes, if the warrior queen would
have just been quiet, it

218
00:10:25,160 --> 00:10:26,650
would've been no
problem at all.

219
00:10:26,650 --> 00:10:28,410
AUDIENCE: If they read
the same thing.

220
00:10:28,410 --> 00:10:29,530
PROFESSOR: If they read
the same thing.

221
00:10:29,530 --> 00:10:30,320
No, that's --

222
00:10:30,320 --> 00:10:34,510
that's probably true, but not
a deeply interesting true.

223
00:10:39,130 --> 00:10:41,920
Well, for instance if she was
male, it would be easier to

224
00:10:41,920 --> 00:10:43,130
segregate the two voices.

225
00:10:43,130 --> 00:10:45,770
If we moved them apart further
it would be easier to

226
00:10:45,770 --> 00:10:47,280
segregate the two voices.

227
00:10:47,280 --> 00:10:48,090
AUDIENCE: If one was singing.

228
00:10:48,090 --> 00:10:49,850
PROFESSOR: If one of them was
singing it would've actually

229
00:10:49,850 --> 00:10:51,710
probably been easier
to segregate them.

230
00:10:51,710 --> 00:10:54,980
So if you change the sort of low
level sensory information,

231
00:10:54,980 --> 00:10:57,270
it would be easier for you to
decide which one to pay

232
00:10:57,270 --> 00:10:58,630
attention to.

233
00:10:58,630 --> 00:11:00,990
This is something
that happens.

234
00:11:00,990 --> 00:11:04,960
Oh, so if you're sitting there
reading the newspaper while

235
00:11:04,960 --> 00:11:07,770
you're trying to listen to this
lecture, odds are you are

236
00:11:07,770 --> 00:11:10,470
missing one of the
two messages.

237
00:11:10,470 --> 00:11:12,620
It's sort of dealer's
choice there.

238
00:11:12,620 --> 00:11:15,320
But it's also not desperately
polite, in

239
00:11:15,320 --> 00:11:16,460
case anybody was wondering.

240
00:11:16,460 --> 00:11:18,040
If you want to read the paper,
you might as well

241
00:11:18,040 --> 00:11:20,210
go somewhere else.

242
00:11:20,210 --> 00:11:25,530
But this happens in the real
world all the time.

243
00:11:25,530 --> 00:11:29,170
There's a version of it known as
the cocktail party effect.

244
00:11:29,170 --> 00:11:34,050
You go to a party and you're
talking to someone, and you

245
00:11:34,050 --> 00:11:36,650
hear, typically, what?

246
00:11:36,650 --> 00:11:39,180
Like, your name, over there.

247
00:11:39,180 --> 00:11:41,330
So you do this selective
attention thing.

248
00:11:41,330 --> 00:11:43,710
You listen to that
conversation.

249
00:11:43,710 --> 00:11:46,140
You seem to be paying attention
to this guy who's

250
00:11:46,140 --> 00:11:48,680
talking to you, but you're
actually listening over there.

251
00:11:48,680 --> 00:11:52,010
The problem is eventually,
this guy stops talking.

252
00:11:52,010 --> 00:11:55,350
And you realize, oh yeah,
I'm supposed to say

253
00:11:55,350 --> 00:11:57,400
something now, right?

254
00:11:57,400 --> 00:12:00,660
I wonder what we're
talking about.

255
00:12:00,660 --> 00:12:03,910
It can lead to a certain amount
of embarrassment.

256
00:12:03,910 --> 00:12:10,500
Now this happens ubiquitously in
sensory systems and across

257
00:12:10,500 --> 00:12:11,300
sensory systems.

258
00:12:11,300 --> 00:12:16,590
So for example right now, until
I mention it, you are

259
00:12:16,590 --> 00:12:19,650
not particularly aware of
the pressure of your

260
00:12:19,650 --> 00:12:21,950
posterior on the seat.

261
00:12:21,950 --> 00:12:24,600
If I direct your attention
to that, you say, oh,

262
00:12:24,600 --> 00:12:26,810
yeah, there it is.

263
00:12:26,810 --> 00:12:30,010
It was presumably there all
along; I wasn't floating a

264
00:12:30,010 --> 00:12:31,430
moment ago.

265
00:12:31,430 --> 00:12:33,550
But until I direct your
attention to it, it doesn't

266
00:12:33,550 --> 00:12:39,960
rise to the level of current
conscious awareness.

267
00:12:39,960 --> 00:12:42,690
And it shows up in vision,
because the visual world is

268
00:12:42,690 --> 00:12:46,360
far too rich for you to process
everywhere at once.

269
00:12:46,360 --> 00:12:51,400
And that's what makes these sort
of Where's Waldo problems

270
00:12:51,400 --> 00:12:53,690
interesting and fun.

271
00:12:53,690 --> 00:12:57,720
If there was not a bottleneck
like this, Waldo man would not

272
00:12:57,720 --> 00:12:59,200
have gotten rich.

273
00:12:59,200 --> 00:12:59,930
Right?

274
00:12:59,930 --> 00:13:01,550
Yeah, where's Waldo?

275
00:13:01,550 --> 00:13:03,310
There he is.

276
00:13:03,310 --> 00:13:04,760
Big deal.

277
00:13:04,760 --> 00:13:05,590
Have you found him?

278
00:13:05,590 --> 00:13:07,910
AUDIENCE: No.

279
00:13:07,910 --> 00:13:09,680
PROFESSOR: Oh look, I have
a little laser today.

280
00:13:09,680 --> 00:13:11,050
Isn't that nice?

281
00:13:11,050 --> 00:13:12,700
So does it work?

282
00:13:15,430 --> 00:13:16,620
That's Waldo up there.

283
00:13:16,620 --> 00:13:19,510
So now you say, oh, that's
really stupid, because I can't

284
00:13:19,510 --> 00:13:20,880
even see him now --

285
00:13:20,880 --> 00:13:25,370
Oh, and we decided to exploit
the technology by having it on

286
00:13:25,370 --> 00:13:26,910
three screens.

287
00:13:26,910 --> 00:13:30,410
There's no added information
there, it's just it was too

288
00:13:30,410 --> 00:13:32,350
cute not to do it.

289
00:13:32,350 --> 00:13:40,340
But if I say, where is the
elephant spraying a car?

290
00:13:40,340 --> 00:13:41,740
You can find it.

291
00:13:41,740 --> 00:13:43,460
You might have noticed it
before if you had been

292
00:13:43,460 --> 00:13:44,230
scrutinizing it.

293
00:13:44,230 --> 00:13:46,450
It was certainly visible
all along, right?

294
00:13:46,450 --> 00:13:49,600
It wasn't that there was a
black hole here before.

295
00:13:49,600 --> 00:13:55,200
It's just that only when you had
the desire to go in search

296
00:13:55,200 --> 00:13:57,160
for it did you manage to direct
your attention to it in

297
00:13:57,160 --> 00:13:59,080
a way that allowed
you to recognize

298
00:13:59,080 --> 00:14:01,650
these couple of objects.

299
00:14:01,650 --> 00:14:07,820
And it's that ability to
constrict your processing

300
00:14:07,820 --> 00:14:10,390
that's really the focus, at
least of the first part of

301
00:14:10,390 --> 00:14:11,400
today's lecture.

302
00:14:11,400 --> 00:14:17,780
Let me show you the equivalent
of the talking example, but

303
00:14:17,780 --> 00:14:19,160
now switch to reading.

304
00:14:19,160 --> 00:14:24,600
What you want to do here is to
look at the little asterisks.

305
00:14:24,600 --> 00:14:29,690
And I'll put up two streams of
text, columns, one on the

306
00:14:29,690 --> 00:14:31,700
left, one on the right.

307
00:14:31,700 --> 00:14:36,260
Nice and big so that
you can read them.

308
00:14:36,260 --> 00:14:38,420
But what you should
notice is --

309
00:14:38,420 --> 00:14:41,830
keep your eyes moving down from
asterisk to asterisk.

310
00:14:41,830 --> 00:14:43,810
What you should notice is you
can read one or the other; you

311
00:14:43,810 --> 00:14:45,660
just can't read both at the same
time, even though they're

312
00:14:45,660 --> 00:14:47,280
nice and big.

313
00:14:47,280 --> 00:14:48,990
Right?

314
00:14:48,990 --> 00:14:52,510
It just doesn't work.

315
00:14:52,510 --> 00:14:54,780
It's not a visual restriction.

316
00:14:54,780 --> 00:14:56,530
It's a central --

317
00:14:56,530 --> 00:15:00,660
it's a capacity limitation
later on in the system.

318
00:15:00,660 --> 00:15:04,400
So this is by way of an answer
to question one on the

319
00:15:04,400 --> 00:15:11,270
handout: what's the problem
that attention is solving?

320
00:15:11,270 --> 00:15:13,930
Attention is solving this
problem of having

321
00:15:13,930 --> 00:15:16,750
too much going on.

322
00:15:16,750 --> 00:15:19,270
Oh, and attention is
a grab bag term.

323
00:15:19,270 --> 00:15:22,610
I'm going to be talking about
visual selective attention.

324
00:15:22,610 --> 00:15:27,310
Attention isn't one thing, like
my laser pointer here.

325
00:15:27,310 --> 00:15:29,670
There are attentional
mechanisms, selective

326
00:15:29,670 --> 00:15:33,000
mechanisms, all over the place
in the nervous system.

327
00:15:33,000 --> 00:15:37,060
So when you are attending to the
pressure of your posterior

328
00:15:37,060 --> 00:15:40,620
on the seat, you are selecting,
probably using a

329
00:15:40,620 --> 00:15:43,460
different set of neural
circuitry than when you're

330
00:15:43,460 --> 00:15:46,810
selecting one of these words.

331
00:15:46,810 --> 00:15:51,160
It's the same basic idea, but
it's not like there's a single

332
00:15:51,160 --> 00:15:55,910
attention box in your
brain somewhere.

333
00:15:55,910 --> 00:15:58,770
OK.

334
00:15:58,770 --> 00:16:03,090
Some things, as we saw in that
auditory demo, the reading

335
00:16:03,090 --> 00:16:06,130
demo, some things escape
the bottleneck.

336
00:16:06,130 --> 00:16:09,520
Some things can be appreciated
everywhere,

337
00:16:09,520 --> 00:16:13,000
all at the same time.

338
00:16:13,000 --> 00:16:20,540
Well, question two is, what
is that set of things?

339
00:16:20,540 --> 00:16:28,780
And the answer is not babies.

340
00:16:28,780 --> 00:16:36,400
The answer is that there is a
limited set of basic features

341
00:16:36,400 --> 00:16:41,950
that can be processed across
the entire visual

342
00:16:41,950 --> 00:16:43,070
field at one time.

343
00:16:43,070 --> 00:16:45,580
Or, you could do it
in auditory space.

344
00:16:45,580 --> 00:16:48,760
There'd be a set of basic
features in auditory space,

345
00:16:48,760 --> 00:16:51,260
too, that could be processed
at the same time.

346
00:16:51,260 --> 00:16:52,660
But I'm going to stick
with vision.

347
00:16:52,660 --> 00:16:55,870
So all these babies
look alike.

348
00:16:55,870 --> 00:16:59,080
It doesn't take much to figure
out that now there is -- da da

349
00:16:59,080 --> 00:17:01,370
da, where'd Mara go?

350
00:17:01,370 --> 00:17:01,950
Oh, there's Mara.

351
00:17:01,950 --> 00:17:05,310
If the baby turns green, you
do something about it.

352
00:17:05,310 --> 00:17:06,010
Right?

353
00:17:06,010 --> 00:17:07,480
It's a highly salient
stimulus.

354
00:17:07,480 --> 00:17:12,200
Or if the baby's head gets
squashed, you know.

355
00:17:12,200 --> 00:17:17,360
So they're a collection of
simple, basic features, like

356
00:17:17,360 --> 00:17:25,530
color, size, orientation, that
are not bottleneck limited in

357
00:17:25,530 --> 00:17:27,580
the same kind of way.

358
00:17:27,580 --> 00:17:31,060
You can find that if there's a
single red thing in the field,

359
00:17:31,060 --> 00:17:33,890
you can find it anywhere
without having

360
00:17:33,890 --> 00:17:35,460
to go hunting around.

361
00:17:39,390 --> 00:17:42,440
Other things that you might
think would be pretty obvious

362
00:17:42,440 --> 00:17:44,320
are not anywhere near
so obvious.

363
00:17:44,320 --> 00:17:49,190
So as you look around here, you
may notice that most of

364
00:17:49,190 --> 00:17:52,460
these baby heads are upside
down and two of them

365
00:17:52,460 --> 00:17:55,040
are right way up.

366
00:17:55,040 --> 00:17:59,630
But it's not like the
green baby head.

367
00:17:59,630 --> 00:18:04,650
You have to go hunting for
upright versus upside down.

368
00:18:04,650 --> 00:18:06,930
Even though that's a very
salient thing in the real

369
00:18:06,930 --> 00:18:09,410
world, whether or not you're
upright, or whether your

370
00:18:09,410 --> 00:18:13,150
baby's upright or upside down.

371
00:18:13,150 --> 00:18:17,630
So there are about, by last
count -- last count was done

372
00:18:17,630 --> 00:18:19,810
by me, as it turns out --

373
00:18:19,810 --> 00:18:24,570
12 to 18 of these things that
seem to escape the bottleneck.

374
00:18:24,570 --> 00:18:25,820
And that's probably about it.

375
00:18:25,820 --> 00:18:31,670
And they are a bunch of simple
things -- well, seemingly

376
00:18:31,670 --> 00:18:36,490
simple things -- like color,
orientation, and size.

377
00:18:36,490 --> 00:18:38,930
Things that you could imagine,
for instance, the earliest

378
00:18:38,930 --> 00:18:42,510
stages of visual cortical
processing doing.

379
00:18:42,510 --> 00:18:45,320
And then there are some other,
more elaborate things that

380
00:18:45,320 --> 00:18:48,310
also escape this bottleneck.

381
00:18:48,310 --> 00:18:52,020
And they're things like -- well,
if you believe my friend

382
00:18:52,020 --> 00:18:54,600
Chen from China, this would this
would be an example of

383
00:18:54,600 --> 00:18:57,390
the importance of topology.

384
00:18:57,390 --> 00:18:59,810
He thinks that the distinction
here is that this has a hole

385
00:18:59,810 --> 00:19:01,580
and this doesn't have a hole.

386
00:19:01,580 --> 00:19:04,600
The other possibility is that
this has line terminations and

387
00:19:04,600 --> 00:19:06,590
that this doesn't.

388
00:19:06,590 --> 00:19:07,840
These are the sort of
things you can fight

389
00:19:07,840 --> 00:19:08,630
about in this field.

390
00:19:08,630 --> 00:19:11,360
But anyway, it's easy to
find that among that.

391
00:19:11,360 --> 00:19:16,170
Curvy things among straight
things are easy.

392
00:19:16,170 --> 00:19:18,740
Orientation in the third
dimension works.

393
00:19:18,740 --> 00:19:21,630
So that cube is pointing up this
direction; these cubes

394
00:19:21,630 --> 00:19:23,270
are pointing down over here.

395
00:19:23,270 --> 00:19:28,570
That turns out to be easy.

396
00:19:28,570 --> 00:19:31,590
Other examples would
include motion.

397
00:19:31,590 --> 00:19:35,030
Though actually, motion makes
an interesting point.

398
00:19:35,030 --> 00:19:38,300
It's easy to detect the presence
of something, but not

399
00:19:38,300 --> 00:19:39,880
so easy to detect its absence.

400
00:19:39,880 --> 00:19:41,300
So imagine the following.

401
00:19:41,300 --> 00:19:44,190
I didn't make a demo of
this; I could have.

402
00:19:44,190 --> 00:19:46,730
Imagine you're looking at the
ground and there's one little

403
00:19:46,730 --> 00:19:48,680
ant moving around.

404
00:19:48,680 --> 00:19:51,610
He's pretty easy
to find, right?

405
00:19:51,610 --> 00:19:54,010
Because motion is one of these
features that you don't have

406
00:19:54,010 --> 00:19:56,490
to go hunting for; it's
just sort of there.

407
00:19:56,490 --> 00:19:58,440
On the other hand, imagine
you're looking at an ant's

408
00:19:58,440 --> 00:20:01,970
nest, and there's
one dead ant.

409
00:20:01,970 --> 00:20:05,990
How easy is it to find one
ant who's not moving?

410
00:20:05,990 --> 00:20:07,390
Not easy.

411
00:20:07,390 --> 00:20:10,500
So the absence of a feature
can be hard to detect.

412
00:20:10,500 --> 00:20:15,230
The presence of a feature, one
of these 12 to 18 basic

413
00:20:15,230 --> 00:20:17,230
features, can be
easy to detect.

414
00:20:17,230 --> 00:20:22,590
Now, how do you actually go
about establishing that

415
00:20:22,590 --> 00:20:24,790
something is easy to find
or hard to find?

416
00:20:24,790 --> 00:20:26,820
I've been doing this in very
qualitative terms.

417
00:20:26,820 --> 00:20:29,460
But now let me explain
how you actually go

418
00:20:29,460 --> 00:20:30,500
about studying this.

419
00:20:30,500 --> 00:20:33,060
What we would pay you $10
an hour for if you

420
00:20:33,060 --> 00:20:35,280
show up in the lab.

421
00:20:35,280 --> 00:20:39,650
What we would do is show you a
computer screen full of stuff,

422
00:20:39,650 --> 00:20:41,200
and ask you a question.

423
00:20:41,200 --> 00:20:43,800
A simple-minded question like,
on the next one, is there a

424
00:20:43,800 --> 00:20:45,300
tilted line?

425
00:20:45,300 --> 00:20:47,480
And what you would be doing is
sitting there with a couple of

426
00:20:47,480 --> 00:20:49,250
computer keys.

427
00:20:49,250 --> 00:20:52,140
Bang one key if the answer is
no, bang another key if the

428
00:20:52,140 --> 00:20:53,010
answer is yes.

429
00:20:53,010 --> 00:20:55,770
Do it as fast and accurately as
you can, and we're going to

430
00:20:55,770 --> 00:20:57,280
measure your reaction time.

431
00:20:57,280 --> 00:21:00,230
The amount of time from the
onset of the stimulus to the

432
00:21:00,230 --> 00:21:01,980
onset of your response.

433
00:21:01,980 --> 00:21:03,220
How fast can you do it?

434
00:21:03,220 --> 00:21:05,690
Well, I don't have keys for
everybody here, so let's just

435
00:21:05,690 --> 00:21:06,830
do it verbally.

436
00:21:06,830 --> 00:21:12,370
Say yes or no as fast as you can
in response to these guys.

437
00:21:12,370 --> 00:21:14,630
Tell me, is there a tilted
line present?

438
00:21:14,630 --> 00:21:15,150
Ready?

439
00:21:15,150 --> 00:21:17,010
AUDIENCE: Yes.

440
00:21:17,010 --> 00:21:18,280
PROFESSOR: Ready?

441
00:21:18,280 --> 00:21:19,200
AUDIENCE: No.

442
00:21:19,200 --> 00:21:20,400
PROFESSOR: Ready?

443
00:21:20,400 --> 00:21:21,750
AUDIENCE: Yes.

444
00:21:21,750 --> 00:21:23,550
PROFESSOR: Ready?

445
00:21:23,550 --> 00:21:23,820
AUDIENCE: No.

446
00:21:23,820 --> 00:21:26,310
PROFESSOR: OK, that's pretty
straightforward.

447
00:21:26,310 --> 00:21:27,000
What's the next thing?

448
00:21:27,000 --> 00:21:29,600
OK.

449
00:21:29,600 --> 00:21:33,840
What you should have heard is
that your answers were given

450
00:21:33,840 --> 00:21:37,530
crisply, in unison, and it
didn't make any real

451
00:21:37,530 --> 00:21:39,570
difference whether there were
lots of vertical lines on the

452
00:21:39,570 --> 00:21:41,960
screen or a few vertical
lines on the screen.

453
00:21:41,960 --> 00:21:47,390
So if we were to collect real
data and to plot the reaction

454
00:21:47,390 --> 00:21:50,400
time in milliseconds --
thousandths of a second -- as

455
00:21:50,400 --> 00:21:53,490
a function of the set size --
the number of items on the

456
00:21:53,490 --> 00:21:58,720
screen -- what you would get
for any of these 12 to 18

457
00:21:58,720 --> 00:22:00,990
items, if you did the experiment
right, is an

458
00:22:00,990 --> 00:22:03,250
essentially flat line here.

459
00:22:03,250 --> 00:22:05,800
This would be the line for
saying yes, it always turns

460
00:22:05,800 --> 00:22:08,350
out to take a little longer, or
typically turns out to take

461
00:22:08,350 --> 00:22:11,270
a little longer to say no, but
it's not dependent on the

462
00:22:11,270 --> 00:22:13,720
number of items on the screen.

463
00:22:13,720 --> 00:22:18,500
So is there an L, is there a
green thing, is there an X

464
00:22:18,500 --> 00:22:19,560
among these pluses?

465
00:22:19,560 --> 00:22:24,840
All those things would produce
similar looking results where

466
00:22:24,840 --> 00:22:28,970
the slope of this reaction
time by set size function

467
00:22:28,970 --> 00:22:31,330
would be, essentially, 0.

468
00:22:31,330 --> 00:22:33,990
Not all tasks behave that way.

469
00:22:33,990 --> 00:22:37,080
So let's do a different one.

470
00:22:37,080 --> 00:22:41,830
In this case you're looking for
the letter T. It can be

471
00:22:41,830 --> 00:22:47,080
rotated by 90 degrees
left or right, or --

472
00:22:47,080 --> 00:22:48,610
maybe it can also be upside
down; I don't

473
00:22:48,610 --> 00:22:49,470
remember what I put in.

474
00:22:49,470 --> 00:22:53,090
But it may not be an upright
T. But it'll be a T. The

475
00:22:53,090 --> 00:22:55,980
distractor items are all L's.

476
00:22:55,980 --> 00:23:00,110
And I just want you to say
as fast as you can,

477
00:23:00,110 --> 00:23:01,890
is there a T present?

478
00:23:01,890 --> 00:23:04,420
Ready?

479
00:23:04,420 --> 00:23:05,370
AUDIENCE: No.

480
00:23:05,370 --> 00:23:07,150
PROFESSOR: Ready?

481
00:23:07,150 --> 00:23:09,720
AUDIENCE: Yes.

482
00:23:09,720 --> 00:23:12,410
Yes.

483
00:23:12,410 --> 00:23:13,300
[INTERPOSING VOICES]

484
00:23:13,300 --> 00:23:16,790
PROFESSOR: OK, ready?

485
00:23:16,790 --> 00:23:17,920
AUDIENCE: Yes.

486
00:23:17,920 --> 00:23:19,680
PROFESSOR: Ready?

487
00:23:19,680 --> 00:23:22,010
AUDIENCE: Yes.

488
00:23:22,010 --> 00:23:23,250
PROFESSOR: You also
heard the speed -

489
00:23:23,250 --> 00:23:26,090
accuracy tradeoff there.

490
00:23:26,090 --> 00:23:32,850
This is a known phenomenon in
reaction time studies, which

491
00:23:32,850 --> 00:23:35,790
is, one can respond very quickly
if you don't sweat the

492
00:23:35,790 --> 00:23:37,730
accuracy things.

493
00:23:37,730 --> 00:23:39,870
And people do that routinely.

494
00:23:39,870 --> 00:23:42,480
When people do that a lot
in our studies, we

495
00:23:42,480 --> 00:23:45,930
call them bad subjects.

496
00:23:45,930 --> 00:23:49,070
And we don't invite them back.

497
00:23:49,070 --> 00:23:53,870
But what you should have heard
there, and should have felt

498
00:23:53,870 --> 00:23:58,200
yourself, is that the responses
were faster when

499
00:23:58,200 --> 00:24:00,880
there were fewer
items present.

500
00:24:00,880 --> 00:24:05,090
And that the responses of the
group, particularly for these

501
00:24:05,090 --> 00:24:07,650
larger set sizes,
were spread out.

502
00:24:07,650 --> 00:24:08,750
Why were they spread out?

503
00:24:08,750 --> 00:24:12,370
Well, some people got lucky.

504
00:24:12,370 --> 00:24:14,370
This thing came up and their
attention happened to be

505
00:24:14,370 --> 00:24:14,950
around here.

506
00:24:14,950 --> 00:24:18,385
Oh look, there's a T. Some
people were unlucky -- oh dee

507
00:24:18,385 --> 00:24:24,460
do dee dee, oh yeah, there's a
T. And some people were trying

508
00:24:24,460 --> 00:24:28,360
to psych out the professor and
said, there was a yes, there

509
00:24:28,360 --> 00:24:29,650
was a no, there was
another yes.

510
00:24:29,650 --> 00:24:31,530
I know about this: there's
going to be a no.

511
00:24:31,530 --> 00:24:34,480
And they said no without doing
anything so boring as to

512
00:24:34,480 --> 00:24:37,510
actually look at the display.

513
00:24:37,510 --> 00:24:43,000
So what you get for data in an
experiment like this would

514
00:24:43,000 --> 00:24:44,620
look much more like this.

515
00:24:44,620 --> 00:24:48,430
As you increase the set size,
now the reaction time

516
00:24:48,430 --> 00:24:54,740
increases in a fairly linear
kind of a way.

517
00:24:54,740 --> 00:24:59,990
The slope on these
is quite fast.

518
00:24:59,990 --> 00:25:02,940
I mean, this is 20 to 30
milliseconds, thousandths of a

519
00:25:02,940 --> 00:25:05,940
second, for each additional
item to say yes, and about

520
00:25:05,940 --> 00:25:08,840
twice that amount to say no.

521
00:25:08,840 --> 00:25:12,090
Depending on how one exactly
models this, this suggests

522
00:25:12,090 --> 00:25:14,610
that you're running through
20 to 40 of

523
00:25:14,610 --> 00:25:16,830
these letters a second.

524
00:25:16,830 --> 00:25:19,880
So you're going through it
quickly, but you're having to

525
00:25:19,880 --> 00:25:21,060
search now.

526
00:25:21,060 --> 00:25:24,130
It's not simply obvious that
there's a T there; you've got

527
00:25:24,130 --> 00:25:25,740
to go and hunt for it.

528
00:25:25,740 --> 00:25:28,850
Over here you can look for the
5, is another typical sort of

529
00:25:28,850 --> 00:25:39,080
task that would produce
results like that.

530
00:25:39,080 --> 00:25:41,050
I wanted to say one other thing
about that, but now I

531
00:25:41,050 --> 00:25:41,850
don't remember what it was.

532
00:25:41,850 --> 00:25:42,290
Oh yes.

533
00:25:42,290 --> 00:25:45,070
What I wanted to say was that
the speed of this tells you

534
00:25:45,070 --> 00:25:50,750
that you're not looking
at the rate of

535
00:25:50,750 --> 00:25:53,270
fixation on each letter.

536
00:25:53,270 --> 00:25:55,390
If you're doing this in the lab,
you make sure that your

537
00:25:55,390 --> 00:25:58,750
stimuli are big enough that you
don't have to move your

538
00:25:58,750 --> 00:26:00,300
eyes to look at each one.

539
00:26:00,300 --> 00:26:04,770
If you have to move your eyes,
your eyes only move at a rate

540
00:26:04,770 --> 00:26:07,060
of about 4 per second.

541
00:26:07,060 --> 00:26:11,490
And so if you have to fixate
each one of the items before

542
00:26:11,490 --> 00:26:13,530
you can tell if it's a T or an
L -- so if you used little

543
00:26:13,530 --> 00:26:17,500
teeny letters -- this slope
would be more like 250

544
00:26:17,500 --> 00:26:24,440
milliseconds per item, not 40 or
50 or something like that.

545
00:26:24,440 --> 00:26:26,970
Attention can move much more
quickly than the eyes.

546
00:26:26,970 --> 00:26:32,390
One of the things that tells
you is that you can attend

547
00:26:32,390 --> 00:26:33,880
where you're not looking.

548
00:26:33,880 --> 00:26:35,800
Something that basketball
players know very well.

549
00:26:35,800 --> 00:26:37,910
When you hear that a basketball
player has great

550
00:26:37,910 --> 00:26:40,610
peripheral vision, what that
really means is that he can be

551
00:26:40,610 --> 00:26:45,310
looking here and he can be
paying attention to his

552
00:26:45,310 --> 00:26:48,290
teammate over there, and throw
the ball and fake out the

553
00:26:48,290 --> 00:26:48,920
opposition.

554
00:26:48,920 --> 00:26:51,110
Because the usual assumption is
that you're attending where

555
00:26:51,110 --> 00:26:52,470
you're looking.

556
00:26:52,470 --> 00:26:54,370
Most of the time that's true.

557
00:26:54,370 --> 00:26:58,510
But OK, so now I'm looking at
this guy wearing red up there.

558
00:26:58,510 --> 00:27:01,430
And he thinks that I'm
actually paying

559
00:27:01,430 --> 00:27:02,520
attention to him.

560
00:27:02,520 --> 00:27:04,480
But I'm not, actually.

561
00:27:04,480 --> 00:27:06,670
Because of acuity limitations, I
have no idea what I'm paying

562
00:27:06,670 --> 00:27:09,830
attention to here, but I think
it's a woman person, and I

563
00:27:09,830 --> 00:27:11,680
think she just moved.

564
00:27:11,680 --> 00:27:13,110
Oh yeah, look, it is
a woman person.

565
00:27:15,670 --> 00:27:18,810
I can move my attention away
from the point of fixation.

566
00:27:18,810 --> 00:27:21,570
And I can move my attention much
more rapidly than I can

567
00:27:21,570 --> 00:27:25,850
move my eyes.

568
00:27:25,850 --> 00:27:32,810
Now, the find the red thing
among green things is a case

569
00:27:32,810 --> 00:27:36,720
where the property of the target
is one of these basic

570
00:27:36,720 --> 00:27:40,800
features and immediately
gets your attention.

571
00:27:40,800 --> 00:27:45,200
The find the 2 among 5's, or
the T among L's is a case

572
00:27:45,200 --> 00:27:47,970
where everything in the
relevant display is

573
00:27:47,970 --> 00:27:50,140
essentially the same
as far as the early

574
00:27:50,140 --> 00:27:51,430
visual system is concerned.

575
00:27:51,430 --> 00:27:54,580
T's among L's, it's a vertical
and a horizontal line among

576
00:27:54,580 --> 00:27:56,460
other vertical and
horizontal lines.

577
00:27:56,460 --> 00:27:59,770
There's nothing in this early
processing the tells those

578
00:27:59,770 --> 00:28:01,570
apart, it turns out.

579
00:28:01,570 --> 00:28:05,530
Most real world searches
are not like that.

580
00:28:05,530 --> 00:28:10,410
In most real world searches, oh
let's see, what do I feel

581
00:28:10,410 --> 00:28:11,750
like looking for?

582
00:28:11,750 --> 00:28:14,740
I'll look for glasses.

583
00:28:14,740 --> 00:28:19,150
If I'm looking for
eyeglasses --

584
00:28:19,150 --> 00:28:21,520
there are some right there,
and there's some more.

585
00:28:24,900 --> 00:28:28,470
There's no process early in my
visual system, you know, some

586
00:28:28,470 --> 00:28:32,330
huge chunk of cortex devoted to
eyeglass detection. it just

587
00:28:32,330 --> 00:28:33,830
doesn't happen.

588
00:28:33,830 --> 00:28:37,010
At the same time, I don't
search around randomly.

589
00:28:37,010 --> 00:28:39,140
No glasses there, no glasses
there, no glasses there, no

590
00:28:39,140 --> 00:28:39,800
glasses there.

591
00:28:39,800 --> 00:28:42,270
I'm searching in an intelligent
fashion.

592
00:28:42,270 --> 00:28:43,670
Here's how you do that.

593
00:28:43,670 --> 00:28:46,370
Let's do one more
basic search.

594
00:28:46,370 --> 00:28:50,250
What you're looking for here
is a red horizontal line.

595
00:28:50,250 --> 00:28:53,920
Tell me as fast as you can
whether it's present.

596
00:28:53,920 --> 00:28:54,620
AUDIENCE: Yes.

597
00:28:54,620 --> 00:28:57,920
PROFESSOR: Now how you do that
is not by having a chunk of

598
00:28:57,920 --> 00:28:59,730
your brain devoted specifically
to read

599
00:28:59,730 --> 00:29:00,410
horizontals.

600
00:29:00,410 --> 00:29:02,380
Oh, remind me later; I've got
to check whether you still

601
00:29:02,380 --> 00:29:04,370
have a [? McCullough ?] effect,
speaking of red

602
00:29:04,370 --> 00:29:05,100
horizontals.

603
00:29:05,100 --> 00:29:06,840
We'll check that out later.

604
00:29:06,840 --> 00:29:14,150
The way you do that is, you
use those 12 to 18 basic

605
00:29:14,150 --> 00:29:16,480
features to guide your attention
around in an

606
00:29:16,480 --> 00:29:17,660
intelligent fashion.

607
00:29:17,660 --> 00:29:20,840
So if you're looking for red
horizontals, you've got

608
00:29:20,840 --> 00:29:22,610
something that can do red.

609
00:29:22,610 --> 00:29:24,310
You know, give me all
the red things.

610
00:29:24,310 --> 00:29:27,620
You've got something that
can do vertical.

611
00:29:27,620 --> 00:29:29,430
Was I looking for red
horizontals or red verticals?

612
00:29:29,430 --> 00:29:31,010
Well, anyway.

613
00:29:31,010 --> 00:29:32,870
This is a red vertical.

614
00:29:32,870 --> 00:29:34,430
You've got something that
can do vertical.

615
00:29:34,430 --> 00:29:36,900
So I've got the red things, I've
got the vertical things.

616
00:29:36,900 --> 00:29:39,360
I can do that early
on in the system.

617
00:29:39,360 --> 00:29:42,480
All I need is something that
will do something like an

618
00:29:42,480 --> 00:29:44,310
intersection operation.

619
00:29:44,310 --> 00:29:48,550
And if I were to guide to my
attention to the intersection

620
00:29:48,550 --> 00:29:51,240
of the set of all red things
and the set of all vertical

621
00:29:51,240 --> 00:29:53,930
things, that'd be a really good
place to look for red

622
00:29:53,930 --> 00:29:54,670
vertical things.

623
00:29:54,670 --> 00:29:57,260
Oh look, there it is.

624
00:29:57,260 --> 00:30:02,250
So what you've got is a front
end that collects information

625
00:30:02,250 --> 00:30:06,190
that can be used to control this
bottleneck to guide your

626
00:30:06,190 --> 00:30:08,820
attention around, to feed
sensible things to the back

627
00:30:08,820 --> 00:30:15,210
end of the system.

628
00:30:15,210 --> 00:30:17,660
I think that's sort
of pictured there.

629
00:30:17,660 --> 00:30:20,690
And the result is that a search
for something like a

630
00:30:20,690 --> 00:30:23,590
red vertical line, it's not as
easy as finding a red thing

631
00:30:23,590 --> 00:30:26,610
among green things, but
it's pretty easy.

632
00:30:26,610 --> 00:30:31,110
It's easier than finding a 2
among 5's or a T among L's, or

633
00:30:31,110 --> 00:30:31,850
anything like that.

634
00:30:31,850 --> 00:30:35,710
Now this sort of guidance comes
in two different forms.

635
00:30:35,710 --> 00:30:38,070
Or you can think of it as coming
in two different forms.

636
00:30:38,070 --> 00:30:42,710
There's a bottom-up form that's

637
00:30:42,710 --> 00:30:44,650
sort of stimulus driven.

638
00:30:44,650 --> 00:30:48,830
And then there's a top-down
form that's user driven by

639
00:30:48,830 --> 00:30:50,260
your desires.

640
00:30:50,260 --> 00:30:54,070
Let me illustrate that with a
couple more searches for a T.

641
00:30:54,070 --> 00:30:58,470
Tell me as fast as you can
whether or not there's a T in

642
00:30:58,470 --> 00:30:59,710
the next display.

643
00:30:59,710 --> 00:31:02,200
Ready?

644
00:31:02,200 --> 00:31:03,430
AUDIENCE: Yes.

645
00:31:03,430 --> 00:31:04,550
PROFESSOR: That was
pretty crisp.

646
00:31:04,550 --> 00:31:06,760
How did you do it?

647
00:31:06,760 --> 00:31:08,620
Muhmuh.

648
00:31:08,620 --> 00:31:10,660
That's what I thought.

649
00:31:10,660 --> 00:31:15,140
Most people probably found
their attention sort of

650
00:31:15,140 --> 00:31:19,000
automatically grabbed by
this one oddball, which

651
00:31:19,000 --> 00:31:23,200
conveniently enough turned out
to be the T. And so rather

652
00:31:23,200 --> 00:31:26,460
than having to search around,
your attention was grabbed

653
00:31:26,460 --> 00:31:31,830
bottom-up to this item.

654
00:31:31,830 --> 00:31:36,240
Top-down is based on what you
know, or what you've been

655
00:31:36,240 --> 00:31:39,090
told, or instructions
that you've

656
00:31:39,090 --> 00:31:41,300
somehow given to yourself.

657
00:31:41,300 --> 00:31:44,190
So I'm going to tell you, if
there's a T in the next

658
00:31:44,190 --> 00:31:46,900
display, it's red.

659
00:31:46,900 --> 00:31:48,070
What happened out there?

660
00:31:48,070 --> 00:31:50,480
Oh, that was another -- that was
also grabbing attention.

661
00:31:50,480 --> 00:31:53,240
It works in the auditory
domain, too.

662
00:31:53,240 --> 00:31:54,540
If we set off an explosion,

663
00:31:54,540 --> 00:31:57,510
unsurprisingly, you would notice.

664
00:31:57,510 --> 00:31:58,560
All right, you ready?

665
00:31:58,560 --> 00:32:01,796
Is there a T in this
next display?

666
00:32:01,796 --> 00:32:02,780
AUDIENCE: Yes.

667
00:32:02,780 --> 00:32:05,640
PROFESSOR: Whoever said no was
another speed-accuracy

668
00:32:05,640 --> 00:32:10,020
tradeoff, try to smoke out the
professor who had a yes on the

669
00:32:10,020 --> 00:32:12,130
last one and therefore must
have a no on this one.

670
00:32:12,130 --> 00:32:13,820
Look at the display!

671
00:32:13,820 --> 00:32:17,670
Anyway, that's not as easy
as the previous one.

672
00:32:17,670 --> 00:32:23,100
But if you searched around, you
probably noticed, or you

673
00:32:23,100 --> 00:32:26,510
may have noticed, that
you were searching

674
00:32:26,510 --> 00:32:27,530
through the red items.

675
00:32:27,530 --> 00:32:29,060
You're not going to bother
searching through the black

676
00:32:29,060 --> 00:32:31,360
items if you know the T
is going to be red.

677
00:32:31,360 --> 00:32:31,670
Right?

678
00:32:31,670 --> 00:32:40,700
So let us suppose we did an
experiment where the T could

679
00:32:40,700 --> 00:32:43,120
be either black or red.

680
00:32:43,120 --> 00:32:46,160
And I show you a bunch of
displays like this, and I vary

681
00:32:46,160 --> 00:32:51,780
the set size, the number
of items on the screen.

682
00:32:51,780 --> 00:32:53,870
Measure your reaction time.

683
00:32:53,870 --> 00:32:57,550
Let's suppose that the slope
of that function was 30

684
00:32:57,550 --> 00:33:01,330
milliseconds an item.

685
00:33:01,330 --> 00:33:04,550
If that's the case, and half
the items are red in this

686
00:33:04,550 --> 00:33:08,450
display -- or on average half
the items are red -- what's

687
00:33:08,450 --> 00:33:12,850
the slope going to look like
if I tell you that the T is

688
00:33:12,850 --> 00:33:15,158
always red if it's present?

689
00:33:15,158 --> 00:33:17,230
AUDIENCE: [INAUDIBLE]

690
00:33:17,230 --> 00:33:17,980
PROFESSOR: Less steep.

691
00:33:17,980 --> 00:33:18,610
Yeah.

692
00:33:18,610 --> 00:33:20,480
Specifically how less steep?

693
00:33:20,480 --> 00:33:21,580
AUDIENCE: [INAUDIBLE]

694
00:33:21,580 --> 00:33:22,370
PROFESSOR: Very less steep.

695
00:33:22,370 --> 00:33:23,970
That's not specific.

696
00:33:23,970 --> 00:33:24,873
I want a number.

697
00:33:24,873 --> 00:33:25,880
AUDIENCE: 15.

698
00:33:25,880 --> 00:33:26,480
PROFESSOR: 15.

699
00:33:26,480 --> 00:33:28,010
Good number.

700
00:33:28,010 --> 00:33:29,440
Right?

701
00:33:29,440 --> 00:33:35,240
If you can eliminate half the
items, the effective rate of

702
00:33:35,240 --> 00:33:37,260
search is going to be
twice as great.

703
00:33:37,260 --> 00:33:38,960
So the slope will
drop in half.

704
00:33:38,960 --> 00:33:41,310
And that's exactly what you get
in experiments like this.

705
00:33:41,310 --> 00:33:42,780
They work very nicely.

706
00:33:42,780 --> 00:33:46,050
If you have only half the items
on the screen relevant,

707
00:33:46,050 --> 00:33:50,660
subjects behave as though they
are only looking through half

708
00:33:50,660 --> 00:33:52,370
of the items.

709
00:33:52,370 --> 00:33:57,820
So by now I have answered
question two: what escapes the

710
00:33:57,820 --> 00:33:58,910
bottleneck of attention?

711
00:33:58,910 --> 00:34:04,660
Well, there are these 12 to 18
basic properties or features

712
00:34:04,660 --> 00:34:11,170
of the world that seem to
escape the bottleneck.

713
00:34:11,170 --> 00:34:15,100
We can study this by measuring
reaction time.

714
00:34:15,100 --> 00:34:17,825
There are other methods, too,
of course, but I was telling

715
00:34:17,825 --> 00:34:21,740
you about the reaction
time methods.

716
00:34:21,740 --> 00:34:23,650
Oh, I see I put Anne Treisman
and feature

717
00:34:23,650 --> 00:34:27,670
integration theory on there.

718
00:34:27,670 --> 00:34:29,750
Don't worry about the feature
integration part.

719
00:34:29,750 --> 00:34:35,450
That's simply to allow me to a
give honor to Anne Treisman

720
00:34:35,450 --> 00:34:39,750
who really founded the modern
study of visual attention,

721
00:34:39,750 --> 00:34:42,700
after having pioneered
an awful lot of

722
00:34:42,700 --> 00:34:45,370
the auditory things.

723
00:34:45,370 --> 00:34:48,480
The auditory demo at the
beginning was a classroom

724
00:34:48,480 --> 00:34:50,390
version of what's called
dichotic listening.

725
00:34:50,390 --> 00:34:54,830
Typically what you do is put on
a pair of headphones, and

726
00:34:54,830 --> 00:34:58,330
you would have one stream of
speech in one ear and one

727
00:34:58,330 --> 00:34:59,990
stream of speech in
the other ear.

728
00:34:59,990 --> 00:35:02,210
And you ask questions about,
if you're attending through

729
00:35:02,210 --> 00:35:06,170
this ear, what can you still
pick up through this ear?

730
00:35:06,170 --> 00:35:09,650
Anne was doing those things in
the late '50s, early '60s.

731
00:35:09,650 --> 00:35:13,800
Went on in the '70s and '80s to
really invent this field of

732
00:35:13,800 --> 00:35:19,600
the study of visual search,
and is still doing great

733
00:35:19,600 --> 00:35:21,870
stuff, now at Princeton.

734
00:35:21,870 --> 00:35:24,630
She was not at Princeton when I
was an undergraduate there,

735
00:35:24,630 --> 00:35:27,800
but she's there now.

736
00:35:27,800 --> 00:35:29,280
All right, so I answered
question three.

737
00:35:29,280 --> 00:35:32,860
And question four I answered
by saying --

738
00:35:32,860 --> 00:35:34,730
oh, conjunction search.

739
00:35:34,730 --> 00:35:38,980
That search for a red vertical
thing is a conjunction of two

740
00:35:38,980 --> 00:35:40,380
basic features.

741
00:35:40,380 --> 00:35:43,180
It's not adequate to know that
it's red; it's not adequate to

742
00:35:43,180 --> 00:35:44,330
know that it's vertical.

743
00:35:44,330 --> 00:35:48,930
The conjunction of those two
sources of information is

744
00:35:48,930 --> 00:35:53,100
adequate, is what defines
the target.

745
00:35:53,100 --> 00:35:57,950
And you can you use this basic
feature information, the basic

746
00:35:57,950 --> 00:36:00,925
attributes of the stimulus, to
guide your attention around in

747
00:36:00,925 --> 00:36:02,590
an intelligent fashion.

748
00:36:02,590 --> 00:36:04,450
So that guidance comes
in two forms.

749
00:36:04,450 --> 00:36:06,170
It can be bottom-up
stimulus driven or

750
00:36:06,170 --> 00:36:10,990
top-down user driven.

751
00:36:10,990 --> 00:36:15,080
All right, so what is that
attention actually doing?

752
00:36:15,080 --> 00:36:23,350
Why is it that you need
to have this --

753
00:36:23,350 --> 00:36:26,990
what is attention making
possible here that wasn't

754
00:36:26,990 --> 00:36:28,110
possible before?

755
00:36:28,110 --> 00:36:30,190
Oh look, it says that
right there.

756
00:36:30,190 --> 00:36:31,860
Or what were those features
doing before

757
00:36:31,860 --> 00:36:33,700
attention shows up?

758
00:36:33,700 --> 00:36:38,060
Well, here is an
answer to that.

759
00:36:38,060 --> 00:36:42,070
The answer is that you've
got all those features.

760
00:36:42,070 --> 00:36:46,250
And in fact, early processes in
the visual system seem to

761
00:36:46,250 --> 00:36:49,510
cut the scene up into what
you might consider to be

762
00:36:49,510 --> 00:36:52,580
proto-objects.

763
00:36:52,580 --> 00:36:55,210
But those features are
just sort of bundled

764
00:36:55,210 --> 00:36:59,840
together with an object.

765
00:36:59,840 --> 00:37:03,970
So before your attention
arrives, something like this

766
00:37:03,970 --> 00:37:08,135
would be red and green and
vertical and horizontal, and

767
00:37:08,135 --> 00:37:11,440
it's got points on
it, or something.

768
00:37:11,440 --> 00:37:17,110
What attention does is to bind
those features together in a

769
00:37:17,110 --> 00:37:21,130
way that makes it possible for
you to know that the greenness

770
00:37:21,130 --> 00:37:23,960
goes with the verticalness here,
and the redness goes

771
00:37:23,960 --> 00:37:25,330
with the horizontalness.

772
00:37:25,330 --> 00:37:27,830
And those points are arranged,
the whole thing's arranged

773
00:37:27,830 --> 00:37:28,450
into a plus.

774
00:37:28,450 --> 00:37:35,500
The argument is that, OK, I
need attention in order to

775
00:37:35,500 --> 00:37:37,670
recognize any given
individual.

776
00:37:37,670 --> 00:37:40,910
Before attention arrives on that
individual, that person

777
00:37:40,910 --> 00:37:44,230
isn't, you know, a black
hole in space.

778
00:37:44,230 --> 00:37:47,400
That person is a loose
bundle of features.

779
00:37:47,400 --> 00:37:50,200
That attention allows me to bind
those features together

780
00:37:50,200 --> 00:37:54,270
in a way that allows me to
understand how they interact,

781
00:37:54,270 --> 00:37:57,090
and what that recognizable
feature might be.

782
00:37:57,090 --> 00:37:58,970
So oh, there's Kristen.

783
00:37:58,970 --> 00:38:01,610
Hey, Kristen, stand
up and wave.

784
00:38:01,610 --> 00:38:03,310
No really, I was plugging
you before.

785
00:38:03,310 --> 00:38:07,770
So if you want to do this for
$10 an hour, go find Kristen.

786
00:38:07,770 --> 00:38:11,500
So all right up, now we can
make fun of Kristen.

787
00:38:11,500 --> 00:38:13,600
So before Kristen arrived --

788
00:38:13,600 --> 00:38:16,400
no, before Kristen arrived,
she was not visible.

789
00:38:16,400 --> 00:38:19,940
Before I attended to Kristen,
there was presumably a

790
00:38:19,940 --> 00:38:24,580
proto-Kristen object out
there that was a

791
00:38:24,580 --> 00:38:27,760
bundle of Kristen bits.

792
00:38:27,760 --> 00:38:30,980
Only when I got my attention
to her -- even though she'd

793
00:38:30,980 --> 00:38:33,360
been visible all along, and I've
looked over there a bunch

794
00:38:33,360 --> 00:38:35,730
of times -- even though she'd
been visible all along, only

795
00:38:35,730 --> 00:38:38,870
when I got my attention to her
could I bind those features

796
00:38:38,870 --> 00:38:44,090
together and make her into
a recognizable Kristen.

797
00:38:44,090 --> 00:38:47,500
Let me see if I can illustrate
that to you with

798
00:38:47,500 --> 00:38:49,540
another demo here.

799
00:38:52,040 --> 00:38:55,000
And the way that's going
to work is --

800
00:38:55,000 --> 00:39:01,290
OK, so what you want to do in
the next slide is to look for

801
00:39:01,290 --> 00:39:04,110
red verticals again.

802
00:39:04,110 --> 00:39:05,620
You ready?

803
00:39:05,620 --> 00:39:10,306
So tell me if you find
a red vertical.

804
00:39:10,306 --> 00:39:11,370
AUDIENCE: Yes.

805
00:39:11,370 --> 00:39:11,780
PROFESSOR: Yeah.

806
00:39:11,780 --> 00:39:15,460
In fact, you might have noticed
there are two of them.

807
00:39:15,460 --> 00:39:16,300
Very easy.

808
00:39:16,300 --> 00:39:17,220
What's the point?

809
00:39:17,220 --> 00:39:19,620
Well, this is a standard guided
search kind of thing.

810
00:39:19,620 --> 00:39:21,620
Give me all the red things;
give me all the vertical

811
00:39:21,620 --> 00:39:23,810
things; look at the intersection
of those two

812
00:39:23,810 --> 00:39:27,330
sets, and oh lookie, there's
two red verticals up there.

813
00:39:27,330 --> 00:39:33,200
Now what I'm going to do is to
simply take the horizontal

814
00:39:33,200 --> 00:39:35,940
here and jump it up to the
middle of the vertical bit

815
00:39:35,940 --> 00:39:37,950
That's why this is in this
sort of odd arrangement.

816
00:39:37,950 --> 00:39:38,970
I'm going to jump it up here.

817
00:39:38,970 --> 00:39:41,560
So I'm going to make a plus,
like those pluses

818
00:39:41,560 --> 00:39:43,860
that we just saw.

819
00:39:43,860 --> 00:39:46,750
The reason for doing this is,
I'm going to keep all the same

820
00:39:46,750 --> 00:39:48,940
pixels on the screen.

821
00:39:48,940 --> 00:39:49,290
Right?

822
00:39:49,290 --> 00:39:51,860
I'm just going to rearrange
where the reds and greens are.

823
00:39:51,860 --> 00:39:54,010
And of course I'm going to
change the location of the red

824
00:39:54,010 --> 00:39:56,580
vertical, because it's really
boring if I keep it in the

825
00:39:56,580 --> 00:39:57,230
same place.

826
00:39:57,230 --> 00:39:59,980
But you're looking for
red vertical again.

827
00:39:59,980 --> 00:40:02,160
Ready?

828
00:40:02,160 --> 00:40:04,400
AUDIENCE: Yes.

829
00:40:04,400 --> 00:40:05,543
PROFESSOR: Who said no?

830
00:40:05,543 --> 00:40:06,800
AUDIENCE: I said woah.

831
00:40:06,800 --> 00:40:07,460
PROFESSOR: Oh, woah.

832
00:40:07,460 --> 00:40:08,690
OK.

833
00:40:08,690 --> 00:40:10,060
Woah's good, woah is good.

834
00:40:12,620 --> 00:40:14,180
Particularly by the time
it says find the two

835
00:40:14,180 --> 00:40:15,690
red vertical lines.

836
00:40:15,690 --> 00:40:20,040
Anyway, you should have
found both of them.

837
00:40:20,040 --> 00:40:23,420
Let's let's check
intuition here.

838
00:40:23,420 --> 00:40:27,110
How many people vote that it
was easier to find the red

839
00:40:27,110 --> 00:40:30,520
verticals when they
were in pluses?

840
00:40:30,520 --> 00:40:31,700
How many vote that it
was easier when

841
00:40:31,700 --> 00:40:33,490
they were ripped apart?

842
00:40:33,490 --> 00:40:36,260
That is the correct intuition.

843
00:40:36,260 --> 00:40:38,000
Actually, I think I
put the data --

844
00:40:38,000 --> 00:40:39,870
I think I realized earlier
than I put half

845
00:40:39,870 --> 00:40:41,210
the data on a slide.

846
00:40:41,210 --> 00:40:43,750
This is the data for looking
for the pluses.

847
00:40:43,750 --> 00:40:47,360
Quite steep slopes of about 50
milliseconds an item here and

848
00:40:47,360 --> 00:40:49,940
about 140 here.

849
00:40:49,940 --> 00:40:54,120
Just looking for the red
verticals when they were in

850
00:40:54,120 --> 00:40:58,110
the disassociated pluses would
have been down here, with a

851
00:40:58,110 --> 00:40:59,950
slope of about 10.

852
00:40:59,950 --> 00:41:03,300
But I somehow left
it off the slide.

853
00:41:03,300 --> 00:41:05,090
Why is this?

854
00:41:05,090 --> 00:41:09,100
Why are the pluses so
much more difficult?

855
00:41:09,100 --> 00:41:13,350
The answer is that before
attention arrives on the

856
00:41:13,350 --> 00:41:18,260
object, these two pluses are
essentially the same thing.

857
00:41:18,260 --> 00:41:21,160
They are red and green and
vertical and horizontal.

858
00:41:21,160 --> 00:41:27,320
And without attention, you just
don't know the difference

859
00:41:27,320 --> 00:41:28,880
between them.

860
00:41:28,880 --> 00:41:34,060
This thing, this square has red
and green and vertical and

861
00:41:34,060 --> 00:41:37,020
horizontal in it, but
it's in two objects.

862
00:41:37,020 --> 00:41:40,800
And so since you direct your
attention to objects -- to

863
00:41:40,800 --> 00:41:45,510
things that are objects; I've
got too many "to"s in there --

864
00:41:45,510 --> 00:41:47,690
this is not a problem
in the way that

865
00:41:47,690 --> 00:41:49,270
these guys are a problem.

866
00:41:49,270 --> 00:41:50,630
In fact anything
that you do --

867
00:41:50,630 --> 00:41:53,380
I don't think I brought the
demo, but anything that you do

868
00:41:53,380 --> 00:41:56,300
to make this less like
a single object

869
00:41:56,300 --> 00:41:57,980
makes the task easier.

870
00:41:57,980 --> 00:42:01,610
So if I was to put a little
shadow on here, so that it

871
00:42:01,610 --> 00:42:04,210
would look like this thing,
the vertical piece was

872
00:42:04,210 --> 00:42:07,270
sticking out in front of
the horizontal piece,

873
00:42:07,270 --> 00:42:09,450
it would get easier.

874
00:42:09,450 --> 00:42:13,090
Because now you could direct
your attention separately to

875
00:42:13,090 --> 00:42:18,040
different planes in depth.

876
00:42:18,040 --> 00:42:22,890
So attention is directed to
objects, and objects are

877
00:42:22,890 --> 00:42:25,160
available ahead of time as
just sort of these loose

878
00:42:25,160 --> 00:42:28,810
configurations, constellations
of features.

879
00:42:28,810 --> 00:42:33,420
Once attention gets there, they
get glued together into

880
00:42:33,420 --> 00:42:34,980
recognizable objects.

881
00:42:34,980 --> 00:42:36,290
All right.

882
00:42:36,290 --> 00:42:43,410
So what happens when you move
away from an attended object?

883
00:42:46,740 --> 00:42:51,350
That's not a unreasonable
question in this framework.

884
00:42:51,350 --> 00:42:52,170
So let's see.

885
00:42:52,170 --> 00:42:52,860
I need --

886
00:42:52,860 --> 00:42:54,010
Rachel.

887
00:42:54,010 --> 00:42:54,970
There's Rachel.

888
00:42:54,970 --> 00:42:55,820
I thought I recognized her.

889
00:42:55,820 --> 00:42:58,050
All right, I have now
recognized Rachel.

890
00:42:58,050 --> 00:42:59,500
Limited number of people
who I actually

891
00:42:59,500 --> 00:43:01,900
recognize by name in here.

892
00:43:01,900 --> 00:43:04,590
And they come to regret it.

893
00:43:04,590 --> 00:43:05,530
But anyway, all right.

894
00:43:05,530 --> 00:43:07,390
So she was here all along.

895
00:43:07,390 --> 00:43:12,060
I happen to have attended to
her and bound Rachel into a

896
00:43:12,060 --> 00:43:13,900
recognizable Rachel object.

897
00:43:13,900 --> 00:43:16,520
I now, without moving
my eyes in fact,

898
00:43:16,520 --> 00:43:18,380
I'm attending elsewhere.

899
00:43:18,380 --> 00:43:21,350
And somebody's up there, again,
my peripheral vision's

900
00:43:21,350 --> 00:43:23,720
lousy, but I can see that
somebody was moving up there.

901
00:43:23,720 --> 00:43:25,980
They waved a piece of white
paper a moment ago.

902
00:43:25,980 --> 00:43:29,700
The question is, when I moved
my attention elsewhere, what

903
00:43:29,700 --> 00:43:31,660
happened to Rachel?

904
00:43:31,660 --> 00:43:34,480
Did she remain bound,
or did she collapse

905
00:43:34,480 --> 00:43:36,545
into Rachel bits again?

906
00:43:36,545 --> 00:43:38,030
AUDIENCE: [INAUDIBLE]

907
00:43:38,030 --> 00:43:40,640
PROFESSOR: What?

908
00:43:40,640 --> 00:43:42,340
She collapsed into
Rachel bits.

909
00:43:42,340 --> 00:43:43,773
How could you tell?

910
00:43:43,773 --> 00:43:45,023
AUDIENCE: [INAUDIBLE]

911
00:43:48,560 --> 00:43:50,400
PROFESSOR: That's why I was
deliberately still looking at

912
00:43:50,400 --> 00:43:52,940
her, to avoid the
issues of blur.

913
00:43:52,940 --> 00:43:57,220
But the way to do this is not to
continue picking on Rachel,

914
00:43:57,220 --> 00:44:02,010
but the switch to dancing
chickens here.

915
00:44:02,010 --> 00:44:04,170
There we have -- you can
tell we're back in

916
00:44:04,170 --> 00:44:05,420
the realm of my artwork.

917
00:44:07,910 --> 00:44:10,800
Oh, I like this, with the
chickens on three screens.

918
00:44:10,800 --> 00:44:12,050
This is so good.

919
00:44:15,470 --> 00:44:17,630
Anyway, I like those a lot.

920
00:44:17,630 --> 00:44:20,910
Now, so you know you know what
you're looking at here.

921
00:44:20,910 --> 00:44:22,840
You're looking at a bunch
of chickens, right?

922
00:44:22,840 --> 00:44:27,500
And they're doing this
little leggy thing.

923
00:44:27,500 --> 00:44:31,000
You would think that, having
recognized that there's a

924
00:44:31,000 --> 00:44:33,390
bunch of chickens there who are
doing this little dance,

925
00:44:33,390 --> 00:44:39,260
that if one of those chickens
fell apart into chicken bits,

926
00:44:39,260 --> 00:44:41,430
that you would notice, right?

927
00:44:41,430 --> 00:44:42,300
Seems reasonable.

928
00:44:42,300 --> 00:44:43,550
How many of you noticed?

929
00:44:46,180 --> 00:44:48,800
Ooh, ooh, very slow
group here.

930
00:44:48,800 --> 00:44:50,230
It should be -- how
many chickens are

931
00:44:50,230 --> 00:44:51,530
there here, about 20?

932
00:44:51,530 --> 00:44:53,930
It should be about one in 20
of you happen to be -- you

933
00:44:53,930 --> 00:44:58,090
have all seen that already.

934
00:45:04,620 --> 00:45:10,140
So one of these chickens
fell apart.

935
00:45:10,140 --> 00:45:12,140
Well, if you think, quite apart
from the fact that the

936
00:45:12,140 --> 00:45:16,410
artwork is a little lame, the
implications are non-lame.

937
00:45:16,410 --> 00:45:20,170
The implication is, all right,
I'm looking at you guys.

938
00:45:20,170 --> 00:45:25,700
I think I'm looking at a bunch
of humanoid life forms.

939
00:45:25,700 --> 00:45:28,480
They're moving a little
bit, stuff like that.

940
00:45:28,480 --> 00:45:32,660
And you would think that if one
of you just went to pieces

941
00:45:32,660 --> 00:45:36,150
here, that I would notice.

942
00:45:36,150 --> 00:45:40,540
The data strongly suggests
that that's not the case.

943
00:45:40,540 --> 00:45:43,170
That I would eventually notice,
as my attention roves

944
00:45:43,170 --> 00:45:46,590
around the room, if it turned
out that, oh my god, not only

945
00:45:46,590 --> 00:45:51,170
has that person not dozed off,
but her head fell off, I would

946
00:45:51,170 --> 00:45:57,880
notice that and react with
according shock and amusement.

947
00:45:57,880 --> 00:46:03,325
The way this experiment is
actually done is not with the

948
00:46:03,325 --> 00:46:04,450
cute little dancing bits.

949
00:46:04,450 --> 00:46:07,620
You'd be looking at a
screen like this,

950
00:46:07,620 --> 00:46:10,940
and you'd hear, beep.

951
00:46:10,940 --> 00:46:13,140
And the question would be, is
there a destroyed chicken?

952
00:46:15,840 --> 00:46:18,130
Yeah, it's there, right?

953
00:46:18,130 --> 00:46:20,280
Beep.

954
00:46:20,280 --> 00:46:20,670
Beep.

955
00:46:20,670 --> 00:46:21,162
AUDIENCE: Yes.

956
00:46:21,162 --> 00:46:22,148
PROFESSOR: Beep.

957
00:46:22,148 --> 00:46:23,134
AUDIENCE: No.

958
00:46:23,134 --> 00:46:23,627
PROFESSOR: Beep.

959
00:46:23,627 --> 00:46:24,120
AUDIENCE: Yes.

960
00:46:24,120 --> 00:46:24,305
PROFESSOR:

961
00:46:24,305 --> 00:46:26,490
And so on.

962
00:46:26,490 --> 00:46:27,070
You can do it.

963
00:46:27,070 --> 00:46:29,570
It's not a difficult task at
all, particularly with a few

964
00:46:29,570 --> 00:46:32,790
big chickens.

965
00:46:32,790 --> 00:46:34,310
But you have to search.

966
00:46:34,310 --> 00:46:37,100
You have to search through
the chickens each time.

967
00:46:37,100 --> 00:46:41,430
And you're no better with a
display that's got the same

968
00:46:41,430 --> 00:46:44,350
fixed number of chickens up
there all the time, compared

969
00:46:44,350 --> 00:46:47,990
to a display which has de novo
chickens popping up out of

970
00:46:47,990 --> 00:46:51,880
nothingness each time.

971
00:46:51,880 --> 00:46:53,970
Oh, the feet are
moving around.

972
00:46:53,970 --> 00:46:56,230
for the demo, why are
the feet doing this

973
00:46:56,230 --> 00:46:58,570
little chicken dance?

974
00:46:58,570 --> 00:47:01,030
Remember I said that motion is
one of these things you can

975
00:47:01,030 --> 00:47:03,000
pick up automatically?

976
00:47:03,000 --> 00:47:07,610
If you don't have something like
the little moving feet,

977
00:47:07,610 --> 00:47:11,500
then when you have a chicken
fall apart - boink - the

978
00:47:11,500 --> 00:47:14,000
movement of the contour,
compared to all the ones that

979
00:47:14,000 --> 00:47:16,610
aren't moving at all
tips you off that

980
00:47:16,610 --> 00:47:17,600
there's something there.

981
00:47:17,600 --> 00:47:20,530
And that tells you that motion's
important, but it

982
00:47:20,530 --> 00:47:23,020
doesn't tell you the interesting
fact that you're

983
00:47:23,020 --> 00:47:28,470
not aware when an otherwise
coherent object falls to bits.

984
00:47:28,470 --> 00:47:33,210
By the way, it turns out you're
also not aware when

985
00:47:33,210 --> 00:47:36,970
previously incoherent material
coheres into a chicken.

986
00:47:36,970 --> 00:47:39,800
We did the classic chicken
soup experiment.

987
00:47:39,800 --> 00:47:44,950
We had a screen full of chicken
bits like this, and

988
00:47:44,950 --> 00:47:47,490
you heard beep, and you had to
figure out whether or not

989
00:47:47,490 --> 00:47:48,680
there was now a chicken
present.

990
00:47:48,680 --> 00:47:50,460
And you had to search
for that, too.

991
00:47:50,460 --> 00:47:53,600
So chickens emerging from the
chicken soup, which you might

992
00:47:53,600 --> 00:47:55,550
think would be striking,
don't turn out

993
00:47:55,550 --> 00:47:57,510
to be striking either.

994
00:47:57,510 --> 00:47:57,940
All right.

995
00:47:57,940 --> 00:48:02,060
Well the chickens are kind
of ugly and complicated.

996
00:48:02,060 --> 00:48:06,160
How bad is this problem?

997
00:48:06,160 --> 00:48:10,670
So let's get basic here.

998
00:48:10,670 --> 00:48:14,120
No more trying to fool you.

999
00:48:14,120 --> 00:48:15,970
Well, of course I'm trying
to fool you.

1000
00:48:15,970 --> 00:48:19,090
No more dancing around chickens,
and then oh, did you

1001
00:48:19,090 --> 00:48:21,660
see -- after the fact I ask you
whether you saw something

1002
00:48:21,660 --> 00:48:23,730
that fell apart.

1003
00:48:23,730 --> 00:48:26,320
These are what?

1004
00:48:26,320 --> 00:48:28,090
Red and green dots.

1005
00:48:28,090 --> 00:48:32,520
If you weren't sure about that,
it says so at the top.

1006
00:48:32,520 --> 00:48:38,950
All I'm going to do is, I'm
going to cue one dot --

1007
00:48:38,950 --> 00:48:40,790
I don't care about any
of the other dots.

1008
00:48:40,790 --> 00:48:46,770
All I want to know is, did that
one dot change color?

1009
00:48:46,770 --> 00:48:49,770
Say yes or no.

1010
00:48:49,770 --> 00:48:50,180
Whoops.

1011
00:48:50,180 --> 00:48:50,640
Where'd it go?

1012
00:48:50,640 --> 00:48:51,890
AUDIENCE: [INAUDIBLE]

1013
00:48:56,440 --> 00:48:59,230
PROFESSOR: Well, the answer
turns out to be no.

1014
00:49:05,610 --> 00:49:11,450
This is such a great exercise in
applied statistics, right?

1015
00:49:11,450 --> 00:49:14,890
How many -- he can't
really be --

1016
00:49:14,890 --> 00:49:18,780
he said the last one, so no.

1017
00:49:18,780 --> 00:49:19,740
AUDIENCE: Yes.

1018
00:49:19,740 --> 00:49:21,750
PROFESSOR: Oh yeah, but he can't
possibly be doing three

1019
00:49:21,750 --> 00:49:25,290
in a row, right?

1020
00:49:25,290 --> 00:49:27,680
AUDIENCE: No.

1021
00:49:27,680 --> 00:49:29,490
PROFESSOR: That does turn
out to be a no.

1022
00:49:29,490 --> 00:49:32,750
Look, you can hear people
going both ways.

1023
00:49:32,750 --> 00:49:35,820
People are terrible at this.

1024
00:49:35,820 --> 00:49:38,340
They're just barely
above chance.

1025
00:49:38,340 --> 00:49:42,850
And the barely above chance is
consistent with them sort of

1026
00:49:42,850 --> 00:49:47,100
sitting on two or three dots.

1027
00:49:47,100 --> 00:49:48,890
Because you're not just doing
a couple of these, you're

1028
00:49:48,890 --> 00:49:53,060
doing hundreds of these,
for $10 an hour.

1029
00:49:53,060 --> 00:49:58,060
So you can sit on a couple of
them and say, if I get really

1030
00:49:58,060 --> 00:50:00,860
lucky and he cues the one I'm
looking at, I'm going to get

1031
00:50:00,860 --> 00:50:01,630
this right.

1032
00:50:01,630 --> 00:50:04,520
And if he doesn't,
I'm clueless.

1033
00:50:04,520 --> 00:50:05,880
I mean, it's red and green.

1034
00:50:05,880 --> 00:50:08,550
It doesn't get more
basic than that.

1035
00:50:08,550 --> 00:50:08,890
Yup?

1036
00:50:08,890 --> 00:50:10,868
AUDIENCE: I have a question.

1037
00:50:10,868 --> 00:50:12,846
Do people's reaction
times change?

1038
00:50:12,846 --> 00:50:18,523
Because red and green, they have
the same after color, or

1039
00:50:18,523 --> 00:50:19,030
afterimage.

1040
00:50:19,030 --> 00:50:20,666
PROFESSOR: They'd better not
have the same after -- they

1041
00:50:20,666 --> 00:50:21,090
have the opposite.

1042
00:50:21,090 --> 00:50:21,380
Yes.

1043
00:50:21,380 --> 00:50:22,253
AUDIENCE: No, no.

1044
00:50:22,253 --> 00:50:25,373
But the opposite of red is
green, and the opposite of

1045
00:50:25,373 --> 00:50:25,743
green is red.

1046
00:50:25,743 --> 00:50:29,200
So if you do yellow and blue
or something else --?

1047
00:50:29,200 --> 00:50:31,320
PROFESSOR: Well, yellow and blue
are also opposite in the

1048
00:50:31,320 --> 00:50:31,790
same sense.

1049
00:50:31,790 --> 00:50:33,150
But it doesn't matter.

1050
00:50:33,150 --> 00:50:34,490
The color does not matter.

1051
00:50:34,490 --> 00:50:37,880
In fact, we can do another one
with different colors.

1052
00:50:37,880 --> 00:50:40,530
Look at this new.

1053
00:50:40,530 --> 00:50:42,220
More cool colors.

1054
00:50:42,220 --> 00:50:46,460
But maybe I was just
being nasty to you.

1055
00:50:46,460 --> 00:50:49,290
Because there were a lot of
dots up there for you to

1056
00:50:49,290 --> 00:50:50,210
choose among.

1057
00:50:50,210 --> 00:50:52,770
So I'll tell you the
relevant dots.

1058
00:50:52,770 --> 00:50:56,720
What I'm going to do here is
I'll ask you about the color

1059
00:50:56,720 --> 00:50:58,210
of specific dots.

1060
00:50:58,210 --> 00:51:00,150
I won't change them.

1061
00:51:00,150 --> 00:51:01,980
I'll just put them up
there and ask you

1062
00:51:01,980 --> 00:51:03,900
about particular dots.

1063
00:51:03,900 --> 00:51:06,790
And what I want you to do
is tell me the color.

1064
00:51:06,790 --> 00:51:10,880
So if I say, what color is that
dot, the answer is --

1065
00:51:10,880 --> 00:51:11,590
AUDIENCE: Purple.

1066
00:51:11,590 --> 00:51:13,150
PROFESSOR: Good.

1067
00:51:13,150 --> 00:51:16,250
If I happen to cover it up with
a black blob, tell me

1068
00:51:16,250 --> 00:51:18,210
what color it was before
I covered it up.

1069
00:51:18,210 --> 00:51:20,680
OK?

1070
00:51:20,680 --> 00:51:21,570
Ready?

1071
00:51:21,570 --> 00:51:22,650
All right, here we go.

1072
00:51:22,650 --> 00:51:24,540
You'll see how this works.

1073
00:51:24,540 --> 00:51:26,340
Where'd it go?

1074
00:51:26,340 --> 00:51:27,430
There we go.

1075
00:51:27,430 --> 00:51:29,280
AUDIENCE: Red.

1076
00:51:29,280 --> 00:51:30,500
Yellow.

1077
00:51:30,500 --> 00:51:33,540
Blue Green.

1078
00:51:33,540 --> 00:51:34,600
Green.

1079
00:51:34,600 --> 00:51:37,370
PROFESSOR: Good.

1080
00:51:37,370 --> 00:51:38,100
See, you're not --

1081
00:51:38,100 --> 00:51:41,240
I put this in because at this
point you might be sitting

1082
00:51:41,240 --> 00:51:44,770
there saying, I'm so hopeless!

1083
00:51:44,770 --> 00:51:46,610
And I wanted to prove to
you that you're not.

1084
00:51:46,610 --> 00:51:48,730
Well, you are, but not
that hopeless.

1085
00:51:48,730 --> 00:51:49,985
All right, ready?

1086
00:51:49,985 --> 00:51:51,460
AUDIENCE: Purple.

1087
00:51:51,460 --> 00:51:52,580
Red.

1088
00:51:52,580 --> 00:51:53,620
Blue.

1089
00:51:53,620 --> 00:51:54,590
Yellow.

1090
00:51:54,590 --> 00:51:55,550
Red.

1091
00:51:55,550 --> 00:51:56,770
Green.

1092
00:51:56,770 --> 00:51:58,623
Yellow.

1093
00:51:58,623 --> 00:51:59,086
PROFESSOR: Ooh,

1094
00:51:59,086 --> 00:52:00,980
a few people actually got it.

1095
00:52:00,980 --> 00:52:02,860
A bunch of people
did the, urp.

1096
00:52:02,860 --> 00:52:04,730
But yes indeed, that
was yellow.

1097
00:52:04,730 --> 00:52:08,130
It was cued before, so we know
you paid attention to it.

1098
00:52:08,130 --> 00:52:11,170
But it was cued about
five items back.

1099
00:52:11,170 --> 00:52:14,400
And so you'd paid
attention to it.

1100
00:52:14,400 --> 00:52:17,390
It didn't take much binding
to say, that's yellow.

1101
00:52:17,390 --> 00:52:20,750
You'd already done all
the work on it.

1102
00:52:20,750 --> 00:52:23,100
Five blobs later, by the
time your attention is

1103
00:52:23,100 --> 00:52:23,800
somewhere else --

1104
00:52:23,800 --> 00:52:27,230
it wasn't invisible during
that time, right?

1105
00:52:27,230 --> 00:52:28,410
You don't really know
what it is.

1106
00:52:28,410 --> 00:52:29,700
All right, try this.

1107
00:52:29,700 --> 00:52:31,150
AUDIENCE: Red.

1108
00:52:31,150 --> 00:52:32,360
Green.

1109
00:52:32,360 --> 00:52:32,480
Red.

1110
00:52:32,480 --> 00:52:33,490
[MURMURING]

1111
00:52:33,490 --> 00:52:37,980
PROFESSOR: A couple of
people caught on.

1112
00:52:37,980 --> 00:52:39,110
He changed it.

1113
00:52:39,110 --> 00:52:40,270
This is what happened here.

1114
00:52:40,270 --> 00:52:42,170
Whoops, not that way.

1115
00:52:42,170 --> 00:52:43,060
Go back.

1116
00:52:43,060 --> 00:52:43,740
OK.

1117
00:52:43,740 --> 00:52:46,290
So this makes a useful
and important point.

1118
00:52:46,290 --> 00:52:48,520
So, red.

1119
00:52:48,520 --> 00:52:52,150
While your attention was
diverted, I changed the color.

1120
00:52:54,850 --> 00:52:57,140
Why is that important?

1121
00:52:57,140 --> 00:53:02,510
What that tells you, with a very
basic sort of stimulus,

1122
00:53:02,510 --> 00:53:07,290
is that the following ought to
be true: I attend to Rachel, I

1123
00:53:07,290 --> 00:53:08,880
attend away.

1124
00:53:08,880 --> 00:53:10,990
While I've attended
away, Rachel is

1125
00:53:10,990 --> 00:53:13,810
replaced by a kangaroo.

1126
00:53:13,810 --> 00:53:16,760
I am now asked, what
was there?

1127
00:53:16,760 --> 00:53:19,240
I say, you know,
it was Rachel.

1128
00:53:19,240 --> 00:53:22,226
The fact that, even though, you
know, still visible in the

1129
00:53:22,226 --> 00:53:24,680
visual field and everything,
until I attend back, I would

1130
00:53:24,680 --> 00:53:27,710
simply not know that something
had changed there.

1131
00:53:27,710 --> 00:53:30,630
So in fact, if you're
worried that --

1132
00:53:30,630 --> 00:53:33,100
the trick here, obviously, since
there are 300 of you or

1133
00:53:33,100 --> 00:53:37,040
so, you want to convince me that
you're paying attention

1134
00:53:37,040 --> 00:53:41,590
in this class, you draw my
attention early in the class,

1135
00:53:41,590 --> 00:53:44,140
and then you subtly sneak out.

1136
00:53:44,140 --> 00:53:46,680
And presumably I think you're
here attending the whole time.

1137
00:53:46,680 --> 00:53:48,440
Because how often
do I get back to

1138
00:53:48,440 --> 00:53:50,150
each individual person?

1139
00:53:50,150 --> 00:53:52,240
Well, actually, it's
not that good.

1140
00:53:52,240 --> 00:53:55,900
Because at 30 to 40 people per
second, I can get back to you

1141
00:53:55,900 --> 00:53:56,450
pretty quickly.

1142
00:53:56,450 --> 00:53:59,000
So forget it.

1143
00:53:59,000 --> 00:54:03,690
But don't forget the basic
point here, which is that

1144
00:54:03,690 --> 00:54:08,490
you're only aware, you're only
updating your knowledge about

1145
00:54:08,490 --> 00:54:14,870
the world, through this narrow
bottleneck of attention, for

1146
00:54:14,870 --> 00:54:16,260
the current object
of attention.

1147
00:54:16,260 --> 00:54:21,530
Everything else, you're
basically working on your

1148
00:54:21,530 --> 00:54:25,250
hypothesis based on the last
time you checked up on it.

1149
00:54:25,250 --> 00:54:28,060
So here is actually what the
data for an experiment like

1150
00:54:28,060 --> 00:54:32,160
this look like.

1151
00:54:32,160 --> 00:54:35,080
So if you didn't pay attention
to the colored dot, right?

1152
00:54:35,080 --> 00:54:37,000
If I never asked about
it at all.

1153
00:54:37,000 --> 00:54:37,920
Here's chance.

1154
00:54:37,920 --> 00:54:40,230
50% in this particular
experiment.

1155
00:54:40,230 --> 00:54:42,060
Because this is a two
color version of it.

1156
00:54:42,060 --> 00:54:43,430
Is it red or is it green?

1157
00:54:43,430 --> 00:54:45,405
You've got about a 50-50
chance of getting it.

1158
00:54:45,405 --> 00:54:47,160
You do a little bit
better than that.

1159
00:54:47,160 --> 00:54:50,950
If it was recently cued -- if
I just asked you whether it

1160
00:54:50,950 --> 00:54:53,120
was red or green -- you
do pretty well.

1161
00:54:53,120 --> 00:54:57,620
But as soon as it's four items
ago, or eight or 12 ago,

1162
00:54:57,620 --> 00:55:00,110
you're back to being
pretty pathetic.

1163
00:55:00,110 --> 00:55:03,180
So you don't keep a good
record of this.

1164
00:55:03,180 --> 00:55:13,760
You're only updating in the
current object of attention.

1165
00:55:13,760 --> 00:55:16,250
This suggests that your memory
is pretty small here.

1166
00:55:16,250 --> 00:55:18,040
We'll talk about memory more
extensively later.

1167
00:55:18,040 --> 00:55:20,330
But let me illustrate
that your memory is

1168
00:55:20,330 --> 00:55:22,320
actually fairly small.

1169
00:55:22,320 --> 00:55:24,130
Here what we're going to do
is, I want you to remember

1170
00:55:24,130 --> 00:55:25,810
these guys.

1171
00:55:25,810 --> 00:55:27,960
Got them?

1172
00:55:27,960 --> 00:55:30,720
OK, take them away.

1173
00:55:30,720 --> 00:55:31,850
Are these the same?

1174
00:55:31,850 --> 00:55:32,550
AUDIENCE: No.

1175
00:55:32,550 --> 00:55:34,450
PROFESSOR: OK, well your memory
isn't that small.

1176
00:55:34,450 --> 00:55:35,430
That's good.

1177
00:55:35,430 --> 00:55:36,540
How about these guys?

1178
00:55:36,540 --> 00:55:37,290
AUDIENCE: No.

1179
00:55:37,290 --> 00:55:37,770
PROFESSOR: No, no, no.

1180
00:55:37,770 --> 00:55:41,590
This is a new set.

1181
00:55:41,590 --> 00:55:41,860
[LAUGHTER]

1182
00:55:41,860 --> 00:55:42,660
Ready?

1183
00:55:42,660 --> 00:55:44,610
Boink.

1184
00:55:44,610 --> 00:55:46,170
AUDIENCE: Yes.

1185
00:55:46,170 --> 00:55:49,120
PROFESSOR: Whoops.

1186
00:55:49,120 --> 00:55:51,770
Sadly, I can't remember.

1187
00:55:51,770 --> 00:55:53,150
Remember these.

1188
00:55:53,150 --> 00:55:54,890
AUDIENCE: [INAUDIBLE]

1189
00:55:54,890 --> 00:55:57,010
They look the same,
don't they?

1190
00:55:57,010 --> 00:55:57,520
AUDIENCE: Yes.

1191
00:55:57,520 --> 00:55:59,010
PROFESSOR: OK.

1192
00:55:59,010 --> 00:56:00,600
So, well.

1193
00:56:03,140 --> 00:56:04,610
How about these?

1194
00:56:04,610 --> 00:56:04,830
AUDIENCE: No.

1195
00:56:04,830 --> 00:56:09,130
No, this is a new set.

1196
00:56:09,130 --> 00:56:11,800
AUDIENCE: Yes.

1197
00:56:11,800 --> 00:56:13,090
PROFESSOR: Yes, something
changed.

1198
00:56:13,090 --> 00:56:17,740
So this time I transposed
the red and the yellow.

1199
00:56:17,740 --> 00:56:19,120
That's a little more difficult,
because I didn't

1200
00:56:19,120 --> 00:56:22,130
introduce a new color.

1201
00:56:22,130 --> 00:56:25,370
How about this?

1202
00:56:25,370 --> 00:56:27,410
AUDIENCE: Yes.

1203
00:56:27,410 --> 00:56:29,550
Yes.

1204
00:56:29,550 --> 00:56:31,810
PROFESSOR: People aren't
quite sure.

1205
00:56:31,810 --> 00:56:33,990
The answer is that the capacity
of this sort of

1206
00:56:33,990 --> 00:56:35,280
memory is about four.

1207
00:56:37,970 --> 00:56:41,980
And some of you will have gotten
the fact that there was

1208
00:56:41,980 --> 00:56:43,320
another transposition, right?

1209
00:56:43,320 --> 00:56:44,310
Of the yellow and the green?

1210
00:56:44,310 --> 00:56:44,480
Whoops.

1211
00:56:44,480 --> 00:56:45,520
The yellows and the greens.

1212
00:56:45,520 --> 00:56:45,655
Yeah.

1213
00:56:45,655 --> 00:56:51,590
The yellow and green guys are --
whoops! -- switching there.

1214
00:56:51,590 --> 00:56:53,410
Some of you will have gotten and
some of you will have not

1215
00:56:53,410 --> 00:56:55,400
gotten it, because some of you
were sitting on the right four

1216
00:56:55,400 --> 00:56:57,030
and some of you were sitting
on the wrong four.

1217
00:56:57,030 --> 00:56:59,380
But it's only about four.

1218
00:56:59,380 --> 00:56:59,990
Four what?

1219
00:56:59,990 --> 00:57:01,540
It turns out to be
four objects.

1220
00:57:01,540 --> 00:57:02,240
Look at this.

1221
00:57:02,240 --> 00:57:03,710
Tell me if anything changes.

1222
00:57:03,710 --> 00:57:05,530
So here we have at least
color, shape, and

1223
00:57:05,530 --> 00:57:09,020
orientation going on.

1224
00:57:09,020 --> 00:57:09,780
AUDIENCE: Yes.

1225
00:57:09,780 --> 00:57:12,370
PROFESSOR: Yeah, most people
will know here that the red

1226
00:57:12,370 --> 00:57:16,670
thing flipped from pointing
up to pointing down.

1227
00:57:16,670 --> 00:57:19,210
That would seem to suggest that
you can keep track of 12

1228
00:57:19,210 --> 00:57:21,750
things, because there are four
colors, four shapes, and four

1229
00:57:21,750 --> 00:57:23,590
orientations.

1230
00:57:23,590 --> 00:57:26,430
But if I spread those out
across 12 objects,

1231
00:57:26,430 --> 00:57:27,490
you'd be very bad.

1232
00:57:27,490 --> 00:57:30,180
It's that you can keep track
of about four objects.

1233
00:57:30,180 --> 00:57:33,110
You can keep track of multiple
features of each of those

1234
00:57:33,110 --> 00:57:36,050
objects, but it's only about
four objects that you

1235
00:57:36,050 --> 00:57:39,940
can keep track of.

1236
00:57:39,940 --> 00:57:43,050
Now let's see.

1237
00:57:43,050 --> 00:57:45,700
How are we doing in
question land?

1238
00:57:45,700 --> 00:57:49,480
OK, so the answer to question
six, at least to the first

1239
00:57:49,480 --> 00:57:52,400
part about it, is that
the objects don't

1240
00:57:52,400 --> 00:57:53,990
seem to stay bound.

1241
00:57:53,990 --> 00:57:58,240
That you need to continuously
update the visual world in

1242
00:57:58,240 --> 00:58:02,030
order to have some idea of what
its current state is, and

1243
00:58:02,030 --> 00:58:06,440
that you're only updating the
current object of attention.

1244
00:58:06,440 --> 00:58:11,690
After a brief break, we will
establish what the Sistine

1245
00:58:11,690 --> 00:58:14,540
Chapel has to tell us
about that fact.

1246
00:58:14,540 --> 00:58:19,990
But those of you who wish may
study this image for the next

1247
00:58:19,990 --> 00:58:21,010
couple of minutes or so.

1248
00:58:21,010 --> 00:58:23,260
And everybody else can
just sort of stretch.

1249
00:58:23,260 --> 00:58:25,280
And then we'll come back.

1250
00:58:28,620 --> 00:58:30,770
While I apologize to Rachel
for picking on her.

1251
00:58:30,770 --> 00:58:34,360
You're not traumatized
for life or anything?

1252
00:58:34,360 --> 00:58:38,590
OK, good.

1253
00:58:38,590 --> 00:58:39,091
[?

1254
00:58:39,091 --> 00:58:40,597
[CROWD NOISES] ?]

1255
00:59:13,756 --> 00:59:14,260
AUDIENCE:

1256
00:59:14,260 --> 00:59:17,460
Have you seen this video they
have where it's a bunch of

1257
00:59:17,460 --> 00:59:18,070
people bouncing balls
to each other?

1258
00:59:18,070 --> 00:59:19,440
PROFESSOR: Yeah.

1259
00:59:19,440 --> 00:59:24,210
That's now gotten to be so
common that I'm not using it.

1260
00:59:27,460 --> 00:59:34,190
[PRIVATE CONVERSATION]

1261
00:59:34,190 --> 00:59:36,310
AUDIENCE: Do you know
who did that?

1262
00:59:36,310 --> 00:59:37,890
PROFESSOR: Yes, Dan Simons.

1263
00:59:37,890 --> 00:59:43,310
Then at Harvard, now at
University of Illinois.

1264
00:59:45,830 --> 00:59:47,490
I will describe a different
Dan Simons

1265
00:59:47,490 --> 00:59:48,740
experiment in a minute.

1266
00:59:48,740 --> 00:59:50,990
OK, let's get back
together here.

1267
00:59:58,230 --> 01:00:05,010
All right, to briefly review.

1268
01:00:05,010 --> 01:00:09,450
the story I have been developing
thus far is that

1269
01:00:09,450 --> 01:00:14,740
even though you are looking at
this scene from the Sistine

1270
01:00:14,740 --> 01:00:17,790
Chapel, and this is the
expulsion from Eden, there's

1271
01:00:17,790 --> 01:00:21,780
Adam and Eve, and this
very cool snake.

1272
01:00:21,780 --> 01:00:24,180
And there's Adam and Eve getting
chucked out, with the

1273
01:00:24,180 --> 01:00:27,070
angel poking them in the head
and stuff like that.

1274
01:00:27,070 --> 01:00:29,720
Even though you are looking at
this, you know what you're

1275
01:00:29,720 --> 01:00:39,720
looking at, that at any given
moment the only thing that's

1276
01:00:39,720 --> 01:00:43,610
really coming through from the
world to recognition is

1277
01:00:43,610 --> 01:00:46,310
whatever is currently being fed
through the bottleneck,

1278
01:00:46,310 --> 01:00:48,220
the current object
of attention.

1279
01:00:48,220 --> 01:00:54,060
And that maybe three or four
objects, the recent status of

1280
01:00:54,060 --> 01:00:57,000
three or four objects is
currently held in this visual

1281
01:00:57,000 --> 01:00:58,340
short term memory.

1282
01:00:58,340 --> 01:01:03,880
The implication here is that I
could change this scene and

1283
01:01:03,880 --> 01:01:05,190
you wouldn't notice.

1284
01:01:05,190 --> 01:01:08,100
So let's find out.

1285
01:01:08,100 --> 01:01:11,260
What did I change?

1286
01:01:11,260 --> 01:01:13,060
AUDIENCE: [INAUDIBLE]

1287
01:01:13,060 --> 01:01:14,970
PROFESSOR: I need a
hand or two here.

1288
01:01:18,910 --> 01:01:19,850
Yeah, sure, what?

1289
01:01:19,850 --> 01:01:22,110
AUDIENCE: [INAUDIBLE]

1290
01:01:22,110 --> 01:01:23,160
PROFESSOR: Oh, the fig leaf.

1291
01:01:23,160 --> 01:01:24,690
The fig leaf, yes.

1292
01:01:24,690 --> 01:01:27,700
The originator of change
blindness, which is what this

1293
01:01:27,700 --> 01:01:31,580
phenomenon is known as, is Ron
Rensink, now at the University

1294
01:01:31,580 --> 01:01:33,680
of British Columbia.

1295
01:01:33,680 --> 01:01:40,310
And he refers to what he calls
"areas of interest." If you

1296
01:01:40,310 --> 01:01:44,420
change something that people are
paying attention to, they

1297
01:01:44,420 --> 01:01:45,230
notice that.

1298
01:01:45,230 --> 01:01:46,370
But of course I knew that.

1299
01:01:46,370 --> 01:01:50,150
And so how many people picked
up the other three changes?

1300
01:01:50,150 --> 01:01:51,650
AUDIENCE: [INAUDIBLE]

1301
01:01:51,650 --> 01:01:54,470
PROFESSOR: Oh, some.

1302
01:01:54,470 --> 01:01:58,340
We have a few people picked
-- what did you get?

1303
01:01:58,340 --> 01:01:59,400
I can't hear you.

1304
01:01:59,400 --> 01:02:01,030
AUDIENCE: [INAUDIBLE]

1305
01:02:01,030 --> 01:02:03,330
PROFESSOR: The stick thing.

1306
01:02:03,330 --> 01:02:04,760
And what?

1307
01:02:04,760 --> 01:02:05,330
Sorry?

1308
01:02:05,330 --> 01:02:07,840
AUDIENCE: [INAUDIBLE]

1309
01:02:07,840 --> 01:02:09,420
Something showed up at
the top that's funny.

1310
01:02:09,420 --> 01:02:11,850
The stick thing moved, and
something showed up at the top

1311
01:02:11,850 --> 01:02:12,430
that's funny.

1312
01:02:12,430 --> 01:02:16,610
So now with that information,
we can go -- whoops.

1313
01:02:16,610 --> 01:02:18,170
AUDIENCE: Right there.

1314
01:02:20,770 --> 01:02:22,240
PROFESSOR: You got the stick.

1315
01:02:22,240 --> 01:02:25,110
See, the reason for the blank
is the same as the moving

1316
01:02:25,110 --> 01:02:28,520
chicken legs, which is that you
don't want to have motion

1317
01:02:28,520 --> 01:02:30,330
transience giving stuff away.

1318
01:02:30,330 --> 01:02:33,700
But if you have motion
transience --

1319
01:02:33,700 --> 01:02:40,190
do do do do do -- you would
think that if you were in the

1320
01:02:40,190 --> 01:02:43,350
Garden of Eden and the branches
were moving from tree

1321
01:02:43,350 --> 01:02:49,230
to tree, or for that matter
Eve's foot was moving to

1322
01:02:49,230 --> 01:02:52,660
Adam's body, you would notice.

1323
01:02:52,660 --> 01:03:00,170
But if you're not attending
to it, you don't notice.

1324
01:03:00,170 --> 01:03:06,430
So this is part of a large set
of phenomena that come under

1325
01:03:06,430 --> 01:03:08,720
the general heading of
change blindness.

1326
01:03:08,720 --> 01:03:11,330
At the break, somebody was
reminding me of one that you

1327
01:03:11,330 --> 01:03:14,120
may have seen because it's
made it onto Nova

1328
01:03:14,120 --> 01:03:15,660
and things like that.

1329
01:03:15,660 --> 01:03:20,740
Done by Dan Simons, where
you're watching people

1330
01:03:20,740 --> 01:03:24,610
apparently play a weird game of
basketball in front of the

1331
01:03:24,610 --> 01:03:29,530
elevators, it turns out in the
psych department at Harvard.

1332
01:03:29,530 --> 01:03:33,150
And while you're doing that,
a guy in a gorilla suit --

1333
01:03:33,150 --> 01:03:36,515
actually, Stan reminded me,
a woman in a gorilla suit.

1334
01:03:36,515 --> 01:03:39,940
It's hard to tell; she's in a
gorilla suit -- walks in,

1335
01:03:39,940 --> 01:03:45,150
walks into the middle of the
game, waves, walks out.

1336
01:03:45,150 --> 01:03:47,630
And then afterwards you ask --

1337
01:03:47,630 --> 01:03:48,890
oh, and you're doing
a demanding task.

1338
01:03:48,890 --> 01:03:52,160
You're supposed to count how
passes there are, or

1339
01:03:52,160 --> 01:03:53,570
something like that.

1340
01:03:53,570 --> 01:03:56,870
And you're asked, did you
notice the person in the

1341
01:03:56,870 --> 01:03:58,040
gorilla suit?

1342
01:03:58,040 --> 01:03:59,930
Well, first you're asked, did
you notice anything weird?

1343
01:03:59,930 --> 01:04:00,990
Eh, no, very boring.

1344
01:04:00,990 --> 01:04:02,200
Notice the person in
the gorilla suit?

1345
01:04:02,200 --> 01:04:02,790
Yeah, right.

1346
01:04:02,790 --> 01:04:03,840
What person in a gorilla suit?

1347
01:04:03,840 --> 01:04:05,840
Show them the video again.

1348
01:04:05,840 --> 01:04:07,090
Oh my --

1349
01:04:09,820 --> 01:04:13,060
Another great Dan Simons
experiment was done when he

1350
01:04:13,060 --> 01:04:14,660
was at Cornell, actually.

1351
01:04:14,660 --> 01:04:19,790
You're on the street in Ithaca,
New York, and some guy

1352
01:04:19,790 --> 01:04:22,210
walks up to you and asks
you for directions.

1353
01:04:22,210 --> 01:04:24,690
Actually it's Dan Simons walks
up to you and asks you for

1354
01:04:24,690 --> 01:04:25,410
directions.

1355
01:04:25,410 --> 01:04:28,590
And so, since you are a nice
person, you start giving Dan

1356
01:04:28,590 --> 01:04:29,560
directions.

1357
01:04:29,560 --> 01:04:32,650
Now you're standing there on the
street and, who knows why,

1358
01:04:32,650 --> 01:04:35,450
but these two guys with
a door are carry a

1359
01:04:35,450 --> 01:04:36,240
door down the street.

1360
01:04:36,240 --> 01:04:38,710
And they walk between
you and Dan.

1361
01:04:38,710 --> 01:04:41,050
Which is kind of rude.

1362
01:04:41,050 --> 01:04:44,070
And then they're off down
the street somewhere.

1363
01:04:44,070 --> 01:04:48,680
And the question is, do you
continue to give directions

1364
01:04:48,680 --> 01:04:51,940
once you see Dan again?

1365
01:04:51,940 --> 01:04:55,980
Of course, the real question is,
did you notice that when

1366
01:04:55,980 --> 01:05:00,530
the door went by, Dan Simons
ducked down and left with the

1367
01:05:00,530 --> 01:05:07,000
door, and his then-student Dan
Levin popped up in his place?

1368
01:05:07,000 --> 01:05:11,530
And it's a different guy.

1369
01:05:11,530 --> 01:05:17,620
50% of the subjects in this
study kept talking.

1370
01:05:20,640 --> 01:05:24,850
A surprisingly large number of
these, on being debriefed

1371
01:05:24,850 --> 01:05:28,580
later, claimed to have
noticed a change.

1372
01:05:28,580 --> 01:05:30,760
Which is a little
strange, right?

1373
01:05:30,760 --> 01:05:33,780
I'm talking to this guy and the
door, and now I'm talking

1374
01:05:33,780 --> 01:05:35,770
-- there's another guy here,
but what the heck?

1375
01:05:35,770 --> 01:05:40,020
He probably wants the answer
to the same question.

1376
01:05:40,020 --> 01:05:42,050
I don't know what
that's about.

1377
01:05:42,050 --> 01:05:45,930
But the important finding
there is that 50% of the

1378
01:05:45,930 --> 01:05:49,400
people behaved as though they
hadn't noticed the change from

1379
01:05:49,400 --> 01:05:52,900
one person to another, who
they were talking to.

1380
01:05:52,900 --> 01:05:53,990
What's going on here?

1381
01:05:53,990 --> 01:05:55,730
Now people aren't completely
stupid.

1382
01:05:55,730 --> 01:05:58,080
The experiment has not been
done, but we kind of

1383
01:05:58,080 --> 01:06:04,670
absolutely know that if I'm
talking to Dan Simons, short

1384
01:06:04,670 --> 01:06:08,700
white guy, and now the door goes
through, and a tall black

1385
01:06:08,700 --> 01:06:13,720
woman is standing there
-- hm, you know?

1386
01:06:13,720 --> 01:06:17,170
Probably that's, again, the sort
of front-end stuff that

1387
01:06:17,170 --> 01:06:19,080
people tend to pick up on.

1388
01:06:19,080 --> 01:06:24,500
But if what you're doing is, I
don't know this guy, but I've

1389
01:06:24,500 --> 01:06:26,140
got a sort of a model
of this guy.

1390
01:06:26,140 --> 01:06:29,010
I'm talking to kind of a short,
white guy person.

1391
01:06:29,010 --> 01:06:31,030
And da da da, I'm still
talking to a

1392
01:06:31,030 --> 01:06:32,190
short, white guy person.

1393
01:06:32,190 --> 01:06:37,310
It's not the same one,
apparently, but that turns out

1394
01:06:37,310 --> 01:06:38,390
not to be a problem.

1395
01:06:38,390 --> 01:06:51,260
This has given rise to a notion
that perception is what

1396
01:06:51,260 --> 01:06:55,030
Kevin O'Regan has called
a grand illusion.

1397
01:06:55,030 --> 01:07:01,870
That the only thing that you
actually see is the current

1398
01:07:01,870 --> 01:07:04,670
object of attention.

1399
01:07:04,670 --> 01:07:12,110
That I think I'm seeing all of
you, but all I'm really doing

1400
01:07:12,110 --> 01:07:14,240
at the moment is paying
attention to the guy with the

1401
01:07:14,240 --> 01:07:16,340
grey stripe on up there.

1402
01:07:16,340 --> 01:07:17,390
Yeah, there he is.

1403
01:07:17,390 --> 01:07:20,840
And now that he's riveted my
attention by waving at me, the

1404
01:07:20,840 --> 01:07:22,800
rest of you are just
not there.

1405
01:07:22,800 --> 01:07:28,450
You are just some sort of grand
illusion floating around

1406
01:07:28,450 --> 01:07:29,100
in my head.

1407
01:07:29,100 --> 01:07:33,890
Now in some sense,
that's correct.

1408
01:07:33,890 --> 01:07:37,560
That what you are seeing
is a creation --

1409
01:07:37,560 --> 01:07:41,100
the burden of the lecture next
time will be to say that

1410
01:07:41,100 --> 01:07:44,130
you're always seeing a theory
about the world.

1411
01:07:44,130 --> 01:07:47,960
You're not seeing the
world directly.

1412
01:07:47,960 --> 01:07:50,970
You're always making an
interpretation, your best

1413
01:07:50,970 --> 01:07:55,680
guess about what the
stimulus means.

1414
01:07:55,680 --> 01:07:58,703
And all the evidence I've been
showing you for the past hour

1415
01:07:58,703 --> 01:08:02,800
or so suggests that you're
only updating that theory

1416
01:08:02,800 --> 01:08:07,240
through this very narrow
bottleneck.

1417
01:08:07,240 --> 01:08:13,350
So in some sense, you are only
seeing this creation of your

1418
01:08:13,350 --> 01:08:17,880
mind, and the only object that
you are currently updating is

1419
01:08:17,880 --> 01:08:24,490
the one that you are currently
attending to.

1420
01:08:24,490 --> 01:08:30,220
But to call the whole thing an
illusion, it seems to me,

1421
01:08:30,220 --> 01:08:34,110
misses an important aspect
of the experience.

1422
01:08:36,850 --> 01:08:40,880
Let's take a very old example.

1423
01:08:40,880 --> 01:08:46,320
The French philosopher of the,
I'm thinking early 18th

1424
01:08:46,320 --> 01:08:51,260
century, whose name I will
now proceed to misspell.

1425
01:08:58,330 --> 01:08:59,050
Does that look --

1426
01:08:59,050 --> 01:09:00,260
any good philosopher sorts?

1427
01:09:00,260 --> 01:09:00,880
That about right?

1428
01:09:00,880 --> 01:09:03,760
Condillac, I believe is how
you pronounce it properly.

1429
01:09:03,760 --> 01:09:09,390
But anyway, Condillac wrote a
number of very interesting

1430
01:09:09,390 --> 01:09:12,500
things about sensation
and perception.

1431
01:09:12,500 --> 01:09:14,970
He's most famous
for his statue.

1432
01:09:14,970 --> 01:09:22,200
His statue that he proposed
as an entity with

1433
01:09:22,200 --> 01:09:24,240
no senses at all.

1434
01:09:24,240 --> 01:09:27,633
And he asked what would the
mental life of this statue be?

1435
01:09:27,633 --> 01:09:30,330
And argued that, in the absence
of any sensory input,

1436
01:09:30,330 --> 01:09:32,390
there would be no mental life.

1437
01:09:32,390 --> 01:09:34,710
And now, he said, let's imagine
opening up, I think he

1438
01:09:34,710 --> 01:09:37,660
opens up the statue's
nostrils.

1439
01:09:37,660 --> 01:09:41,510
And argues that the entire
mental life of this statue is

1440
01:09:41,510 --> 01:09:43,140
now the smell.

1441
01:09:43,140 --> 01:09:45,180
Whatever, I think he waves
a rose under it or

1442
01:09:45,180 --> 01:09:46,910
something like that.

1443
01:09:46,910 --> 01:09:52,390
But a little further on he has
a different example where he

1444
01:09:52,390 --> 01:09:57,680
says imagine, you're in a dark
-- a dark chateau, I believe.

1445
01:09:57,680 --> 01:10:00,630
And it's completely pitch
black, because

1446
01:10:00,630 --> 01:10:01,810
of these heavy curtains.

1447
01:10:01,810 --> 01:10:06,550
And it's morning, and you
throw open the curtains.

1448
01:10:06,550 --> 01:10:09,020
If it were the case -- this is
not what he's saying, but if

1449
01:10:09,020 --> 01:10:11,770
it were the case that all of
vision was nothing but a grand

1450
01:10:11,770 --> 01:10:14,430
illusion, you only saw the
spotlight of attention, this

1451
01:10:14,430 --> 01:10:16,080
one thing that you're attending
to at any one

1452
01:10:16,080 --> 01:10:21,220
moment, your experience of this
brand new scene ought to

1453
01:10:21,220 --> 01:10:27,000
be like sort of a weird
paint brush.

1454
01:10:27,000 --> 01:10:29,170
Initially, I don't
see nothin'.

1455
01:10:29,170 --> 01:10:31,100
Because I haven't attended
to anything.

1456
01:10:31,100 --> 01:10:32,940
Now I attend to an object.

1457
01:10:32,940 --> 01:10:35,660
And now this person,
object, is the only

1458
01:10:35,660 --> 01:10:36,300
thing in the scene.

1459
01:10:36,300 --> 01:10:37,110
And boom, boom, boom.

1460
01:10:37,110 --> 01:10:39,110
And I slowly fill you in.

1461
01:10:39,110 --> 01:10:41,800
That's not the impression
you get ever when

1462
01:10:41,800 --> 01:10:42,780
you see a new scene.

1463
01:10:42,780 --> 01:10:45,140
You may not know what you're
looking at, but you see

1464
01:10:45,140 --> 01:10:48,310
something everywhere
instantly.

1465
01:10:48,310 --> 01:10:52,390
And the grand illusion thing
misses the fact that you're

1466
01:10:52,390 --> 01:10:58,030
somehow sensing something about
the entire visual field

1467
01:10:58,030 --> 01:10:59,110
all at once.

1468
01:10:59,110 --> 01:11:06,080
Let me offer a way of
understanding that that will

1469
01:11:06,080 --> 01:11:09,890
then tie back to the visual
physiology that I was talking

1470
01:11:09,890 --> 01:11:11,740
about in the last lecture.

1471
01:11:11,740 --> 01:11:13,270
Here's the idea.

1472
01:11:13,270 --> 01:11:22,730
Early in your visual system,
you've got the processes that,

1473
01:11:22,730 --> 01:11:25,670
sort of a big river of
information that tells you

1474
01:11:25,670 --> 01:11:31,740
about those 12 to 18 features or
attributes that you can get

1475
01:11:31,740 --> 01:11:33,490
out -- these are eyes.

1476
01:11:33,490 --> 01:11:35,230
This is my drawing again.

1477
01:11:35,230 --> 01:11:38,350
So from your eyes, you've got
this big flow of information

1478
01:11:38,350 --> 01:11:40,160
up into your brain.

1479
01:11:40,160 --> 01:11:45,840
And at some point, it hits this
bottleneck that's taken

1480
01:11:45,840 --> 01:11:48,290
care of by attention.

1481
01:11:48,290 --> 01:11:51,850
Object recognition, the ability
to tell that that's a

1482
01:11:51,850 --> 01:11:57,240
branch, that that's a snake, and
so on, only one object at

1483
01:11:57,240 --> 01:12:03,960
a time can go in and come out
and rise to the level of some

1484
01:12:03,960 --> 01:12:11,210
sort of perceptual awareness,
populating your visual

1485
01:12:11,210 --> 01:12:12,160
experience.

1486
01:12:12,160 --> 01:12:17,620
And that bottleneck is guided
by these collection of basic

1487
01:12:17,620 --> 01:12:19,610
features that you've got.

1488
01:12:19,610 --> 01:12:23,060
If you know you're looking for
red stuff, you set these

1489
01:12:23,060 --> 01:12:24,210
settings for red.

1490
01:12:24,210 --> 01:12:31,400
And maybe vertical, and big
and moving and so on.

1491
01:12:31,400 --> 01:12:34,940
And so you can regulate what
gets through here.

1492
01:12:34,940 --> 01:12:37,510
And only the one thing
at any one time is

1493
01:12:37,510 --> 01:12:39,950
getting up into there.

1494
01:12:39,950 --> 01:12:47,350
And so the current object of
attention gets to rise to

1495
01:12:47,350 --> 01:12:52,630
awareness, and you know what
you're looking at.

1496
01:12:52,630 --> 01:12:55,950
That's the story that I've
told you to this point.

1497
01:12:55,950 --> 01:12:57,930
That's the story that gives
rise to the notion that

1498
01:12:57,930 --> 01:12:59,390
everything else in the
visual field is

1499
01:12:59,390 --> 01:13:00,900
some sort of an illusion.

1500
01:13:00,900 --> 01:13:04,720
But look, when I was doing that
red and green dot thing,

1501
01:13:04,720 --> 01:13:07,180
it wasn't that you didn't see
the other red and green dots.

1502
01:13:07,180 --> 01:13:07,840
They were there.

1503
01:13:07,840 --> 01:13:11,510
You just somehow had a very
impoverished ability to tell

1504
01:13:11,510 --> 01:13:14,250
me anything about them.

1505
01:13:14,250 --> 01:13:17,690
And a way to think about that
is to propose that there's

1506
01:13:17,690 --> 01:13:23,250
another pathway, another big
fat river of information

1507
01:13:23,250 --> 01:13:27,520
about, say, these 12 to 18
attributes, that isn't limited

1508
01:13:27,520 --> 01:13:28,780
by the bottleneck.

1509
01:13:28,780 --> 01:13:31,320
But that it doesn't let you --

1510
01:13:31,320 --> 01:13:32,780
it's not a cheat.

1511
01:13:32,780 --> 01:13:35,450
This doesn't now let you go
and recognize objects

1512
01:13:35,450 --> 01:13:36,850
everywhere all at once.

1513
01:13:36,850 --> 01:13:39,540
It can only do a few things.

1514
01:13:39,540 --> 01:13:42,190
It can sort of give you the
statistics of the world.

1515
01:13:42,190 --> 01:13:44,880
You know, I'm looking out at you
guys and I'm seeing a sort

1516
01:13:44,880 --> 01:13:49,050
of texture of people
amongst purple.

1517
01:13:49,050 --> 01:13:57,970
And that sort of impression of
purpleness, of a tilted plane,

1518
01:13:57,970 --> 01:14:01,700
is the sort of thing that you
might get out of this big,

1519
01:14:01,700 --> 01:14:05,660
broad, unrestricted,
nonselective, as it's labeled

1520
01:14:05,660 --> 01:14:07,660
on there, pathway.

1521
01:14:07,660 --> 01:14:11,460
There's evidence that you can
get a little bit of semantic

1522
01:14:11,460 --> 01:14:12,000
information.

1523
01:14:12,000 --> 01:14:15,160
Semantic means the meaning,
when you're talking about

1524
01:14:15,160 --> 01:14:17,590
language, it's the meaning of
the utterance, let's say.

1525
01:14:17,590 --> 01:14:19,710
When you're talking about
vision, it's the meaning of

1526
01:14:19,710 --> 01:14:22,130
the stimulus.

1527
01:14:22,130 --> 01:14:27,420
So I might get the notion that
I'm in an enclosed space.

1528
01:14:27,420 --> 01:14:30,470
This pathway by itself is not
going to tell me what enclosed

1529
01:14:30,470 --> 01:14:31,070
space I'm in.

1530
01:14:31,070 --> 01:14:32,270
But I'm in a space.

1531
01:14:32,270 --> 01:14:34,660
There's a tilted
surface there.

1532
01:14:34,660 --> 01:14:35,700
And so on.

1533
01:14:35,700 --> 01:14:39,090
But this is going to give me,
that broad pathway is going to

1534
01:14:39,090 --> 01:14:41,370
give me the feeling that there's
something happening

1535
01:14:41,370 --> 01:14:43,220
everywhere.

1536
01:14:43,220 --> 01:14:47,240
And this pathway is going to
tell me what's happening

1537
01:14:47,240 --> 01:14:49,900
specifically here, now.

1538
01:14:49,900 --> 01:14:53,540
And between the two of them, I
can build up an idea in my

1539
01:14:53,540 --> 01:14:57,190
head of, oh, I'm in 10-250.

1540
01:14:57,190 --> 01:14:59,360
I'm talking to this bunch of
people, some of whom I know by

1541
01:14:59,360 --> 01:15:02,440
name, some of whom I recognize
because they've been here

1542
01:15:02,440 --> 01:15:04,550
before, and so on.

1543
01:15:04,550 --> 01:15:08,640
And I can keep updating that 20,
30 times a second through

1544
01:15:08,640 --> 01:15:09,620
this pathway.

1545
01:15:09,620 --> 01:15:13,020
And I can keep experiencing
something, that sort of

1546
01:15:13,020 --> 01:15:17,290
wallpaper of the world effect,
through this other pathway.

1547
01:15:17,290 --> 01:15:21,680
Now that ties back, it might
tie back to things that we

1548
01:15:21,680 --> 01:15:23,360
talked about before.

1549
01:15:23,360 --> 01:15:28,020
If you remember the idea that
you can broadly cut visual

1550
01:15:28,020 --> 01:15:31,760
processing, visual cortical
processing, into two big

1551
01:15:31,760 --> 01:15:34,330
pathways, a what and
a where pathway.

1552
01:15:34,330 --> 01:15:38,650
A what pathway going down into
the temporal lobe, and a where

1553
01:15:38,650 --> 01:15:43,110
pathway going up into
the parietal lobe.

1554
01:15:43,110 --> 01:15:46,070
This selective pathway, this
thing that only does one

1555
01:15:46,070 --> 01:15:48,980
object at a time, would
then be mapped

1556
01:15:48,980 --> 01:15:51,460
onto the what pathway.

1557
01:15:51,460 --> 01:15:55,740
What am I looking at, what am
I attending to right now?

1558
01:15:55,740 --> 01:16:01,350
If you were to lesion that, if
you were to lesion it, or you

1559
01:16:01,350 --> 01:16:05,540
were to have damage to the
temporal lobe of your brain,

1560
01:16:05,540 --> 01:16:08,310
you might well end up
with an agnosia.

1561
01:16:08,310 --> 01:16:11,050
That's not a term that ended up
on the handout, so you want

1562
01:16:11,050 --> 01:16:12,450
to write that one down.

1563
01:16:15,620 --> 01:16:20,470
An agnosia is a failure
to know, if you like.

1564
01:16:20,470 --> 01:16:22,060
To know what something is.

1565
01:16:22,060 --> 01:16:25,840
So an agnosic, if you have a
person with a fairly global

1566
01:16:25,840 --> 01:16:30,310
agnosia, visual agnosia, they
would be able to say, yeah,

1567
01:16:30,310 --> 01:16:32,420
I'm looking at a bunch of
objects here, but I don't know

1568
01:16:32,420 --> 01:16:34,190
what they are.

1569
01:16:34,190 --> 01:16:35,940
Here's this object.

1570
01:16:35,940 --> 01:16:37,670
It's sort of orange.

1571
01:16:37,670 --> 01:16:41,420
It's got orange and brown
and white blobs on it.

1572
01:16:41,420 --> 01:16:44,410
And it's got this very long
part, and there are these four

1573
01:16:44,410 --> 01:16:47,630
pointy things coming off
the bottom of it.

1574
01:16:47,630 --> 01:16:49,640
I've got no idea what
that is; maybe it's

1575
01:16:49,640 --> 01:16:52,140
furniture of some sort.

1576
01:16:52,140 --> 01:16:54,120
You'd look at it and say,
that's a giraffe.

1577
01:16:54,120 --> 01:16:57,400
An agnosic would be able to
tell you about it, but not

1578
01:16:57,400 --> 01:16:59,770
know that it was a giraffe.

1579
01:16:59,770 --> 01:17:04,070
Smaller lesions produce rather
specific agnosias.

1580
01:17:04,070 --> 01:17:06,540
There are reports in the
literature of agnosias

1581
01:17:06,540 --> 01:17:12,050
specific to, say, fruits
and vegetables.

1582
01:17:12,050 --> 01:17:23,590
More common is a form of agnosia
called prosopagnosia,

1583
01:17:23,590 --> 01:17:27,210
which is a specific inability
to recognize faces.

1584
01:17:27,210 --> 01:17:29,480
You know that it's a face, it's
got two eyes, it's got a

1585
01:17:29,480 --> 01:17:29,935
nose and mouth.

1586
01:17:29,935 --> 01:17:32,420
You don't know who it is.

1587
01:17:32,420 --> 01:17:37,740
Small lesions down in that
pathway can produce

1588
01:17:37,740 --> 01:17:39,130
that sort of damage.

1589
01:17:39,130 --> 01:17:42,360
That would suggest, then, that
the other pathway ought to be

1590
01:17:42,360 --> 01:17:46,000
mapped onto the where pathway.

1591
01:17:46,000 --> 01:17:53,020
And if you get bilateral damage,
for instance, to the

1592
01:17:53,020 --> 01:17:57,670
parietal lobe, you can end up
with a disorder known as

1593
01:17:57,670 --> 01:18:01,500
Balint's syndrome -- might as
well write the word down here.

1594
01:18:01,500 --> 01:18:03,630
Named after Balint --

1595
01:18:06,360 --> 01:18:10,530
that has as one of its
properties what's called a

1596
01:18:10,530 --> 01:18:12,040
simultagnosia.

1597
01:18:12,040 --> 01:18:16,060
This is a situation where you
can recognize an object if you

1598
01:18:16,060 --> 01:18:17,490
can get your attention on it.

1599
01:18:17,490 --> 01:18:24,140
But that's the only thing you
can respond to, in some sense.

1600
01:18:24,140 --> 01:18:28,030
It is as if the grand illusion
theory was really right, that

1601
01:18:28,030 --> 01:18:30,650
you can only see the current
object of attention.

1602
01:18:30,650 --> 01:18:34,650
So you do something like this
with a simultagnosic, say,

1603
01:18:34,650 --> 01:18:36,730
what's that?

1604
01:18:36,730 --> 01:18:37,800
Draw his attention to it.

1605
01:18:37,800 --> 01:18:39,690
That's a book.

1606
01:18:39,690 --> 01:18:42,810
OK, what else have
we got here?

1607
01:18:42,810 --> 01:18:44,290
OK, what's that?

1608
01:18:44,290 --> 01:18:45,890
That's a cell phone.

1609
01:18:45,890 --> 01:18:46,580
What's that?

1610
01:18:46,580 --> 01:18:47,800
That's a cell phone.

1611
01:18:47,800 --> 01:18:48,400
Anything else?

1612
01:18:48,400 --> 01:18:49,620
No.

1613
01:18:49,620 --> 01:18:50,200
What's that?

1614
01:18:50,200 --> 01:18:51,410
That's a book.

1615
01:18:51,410 --> 01:18:52,060
What's that?

1616
01:18:52,060 --> 01:18:53,350
That's a book.

1617
01:18:53,350 --> 01:18:53,970
Anything else?

1618
01:18:53,970 --> 01:18:54,880
No.

1619
01:18:54,880 --> 01:18:58,510
So one object at a time.

1620
01:18:58,510 --> 01:19:01,960
As if the where of the world
had disappeared.

1621
01:19:01,960 --> 01:19:04,680
If you get damage -- we'll talk
about this more later in

1622
01:19:04,680 --> 01:19:08,500
the term -- but if you get
damage to the parietal lobe on

1623
01:19:08,500 --> 01:19:12,220
one side, particularly on the
right side, what you can end

1624
01:19:12,220 --> 01:19:15,980
up with is a disorder
known as neglect.

1625
01:19:15,980 --> 01:19:22,370
It comes in a variety of
flavors, again depending on

1626
01:19:22,370 --> 01:19:23,350
the particular lesion.

1627
01:19:23,350 --> 01:19:28,170
But the characteristic is, you
ignore the contralateral, the

1628
01:19:28,170 --> 01:19:29,690
other side.

1629
01:19:29,690 --> 01:19:32,650
Now that can be the other side
of space, so that if I'm a

1630
01:19:32,650 --> 01:19:36,240
patient with a right hemisphere
parietal lesion and

1631
01:19:36,240 --> 01:19:40,750
I'm looking at MIT volleyball
here, everything in the left

1632
01:19:40,750 --> 01:19:44,470
visual field, I would
simply ignore.

1633
01:19:44,470 --> 01:19:47,220
I would behave as though
it did not exist.

1634
01:19:47,220 --> 01:19:50,300
If I took away everything else
and put a stimulus in my left

1635
01:19:50,300 --> 01:19:52,440
visual field, I could show
that the patient

1636
01:19:52,440 --> 01:19:54,290
could still see it.

1637
01:19:54,290 --> 01:19:59,880
But with a full visual field,
he behaves as though there's

1638
01:19:59,880 --> 01:20:02,990
nothing there at all.

1639
01:20:02,990 --> 01:20:07,360
Patients with neglect will
do weird things, like --

1640
01:20:07,360 --> 01:20:08,870
they're in the hospital,
typically, because

1641
01:20:08,870 --> 01:20:10,290
they've had a stroke.

1642
01:20:10,290 --> 01:20:12,000
You give them their dinner.

1643
01:20:12,000 --> 01:20:15,020
They eat everything on the right
side of the plate and

1644
01:20:15,020 --> 01:20:17,170
leave everything on the left
side of the plate.

1645
01:20:17,170 --> 01:20:17,710
Why?

1646
01:20:17,710 --> 01:20:18,760
Because they didn't like
the mashed potatoes?

1647
01:20:18,760 --> 01:20:18,940
No.

1648
01:20:18,940 --> 01:20:20,510
If you rotate the plate, they'll
eat the stuff on the

1649
01:20:20,510 --> 01:20:22,040
other side of the plate.

1650
01:20:22,040 --> 01:20:27,470
It's as if it just didn't
exist, in some fashion.

1651
01:20:27,470 --> 01:20:30,020
Now, you'll remember the
parietal lobe is also where

1652
01:20:30,020 --> 01:20:31,990
you get the representation
of the body surface,

1653
01:20:31,990 --> 01:20:33,030
and stuff like that.

1654
01:20:33,030 --> 01:20:36,570
So neglect patients can also
be patients who neglect one

1655
01:20:36,570 --> 01:20:39,130
half of their body, and
deny that part of

1656
01:20:39,130 --> 01:20:41,530
their body is theirs.

1657
01:20:41,530 --> 01:20:43,780
This is a little easier to
understand if you figure that

1658
01:20:43,780 --> 01:20:46,960
the stroke might well have also
knocked out the ability

1659
01:20:46,960 --> 01:20:48,530
to control that side
of your body.

1660
01:20:48,530 --> 01:20:51,060
So a stroke on the right
might leave you

1661
01:20:51,060 --> 01:20:53,060
paralyzed on the left.

1662
01:20:53,060 --> 01:20:57,200
But you can end up with
situations like one described,

1663
01:20:57,200 --> 01:21:00,650
I think, by Oliver Sacks in
one of his books, where a

1664
01:21:00,650 --> 01:21:05,010
patient is saying, "This
is a cheap hospital.

1665
01:21:05,010 --> 01:21:07,570
This is a really cheap, lousy
hospital." How do you know

1666
01:21:07,570 --> 01:21:10,820
it's a cheap, lousy hospital?
"Because they're doubling up

1667
01:21:10,820 --> 01:21:14,360
on beds." What you mean they're
doubling up on beds?

1668
01:21:14,360 --> 01:21:18,270
He says, "Look at that leg.

1669
01:21:18,270 --> 01:21:23,390
That's not my leg." So you can
get, this is somebody looking

1670
01:21:23,390 --> 01:21:29,860
at their own leg and denying
that that leg belongs to them.

1671
01:21:29,860 --> 01:21:36,080
That's another aspect
of neglect.

1672
01:21:36,080 --> 01:21:42,430
OK, what I'm going to do next
time is to talk about the way

1673
01:21:42,430 --> 01:21:46,050
in which you make hypotheses
about the world.