1
00:00:00,000 --> 00:00:00,080

2
00:00:00,080 --> 00:00:01,670
The following
content is provided

3
00:00:01,670 --> 00:00:03,820
under a Creative
Commons license.

4
00:00:03,820 --> 00:00:06,550
Your support will help MIT
OpenCourseWare continue

5
00:00:06,550 --> 00:00:10,160
to offer high quality
educational resources for free.

6
00:00:10,160 --> 00:00:12,700
To make a donation or to
view additional materials

7
00:00:12,700 --> 00:00:16,620
from hundreds of MIT courses,
visit MIT OpenCourseWare

8
00:00:16,620 --> 00:00:17,275
at ocw.mit.edu.

9
00:00:17,275 --> 00:00:25,720

10
00:00:25,720 --> 00:00:28,360
PROFESSOR: All right, so
this brief reminder here

11
00:00:28,360 --> 00:00:34,990
is about the basic requirements
in this course, which

12
00:00:34,990 --> 00:00:40,300
are that we have
two written reports.

13
00:00:40,300 --> 00:00:43,930
And that's the main thing I want
to remind you of, one in vision

14
00:00:43,930 --> 00:00:45,670
and one in audition.

15
00:00:45,670 --> 00:00:50,490
And I presume all of you
have the syllabus that

16
00:00:50,490 --> 00:00:51,950
was a hand out.

17
00:00:51,950 --> 00:00:54,600
And in that syllabus--
and if you don't have it,

18
00:00:54,600 --> 00:00:57,260
we can provide you, of
course, with a copy.

19
00:00:57,260 --> 00:00:59,910
And on the second
page of the syllabus,

20
00:00:59,910 --> 00:01:03,320
it says written
report, vision part,

21
00:01:03,320 --> 00:01:11,240
and this report is one that is
based on an article published

22
00:01:11,240 --> 00:01:15,540
many years ago, which is
very influential article back

23
00:01:15,540 --> 00:01:21,640
in 1967 about the so-called
accessory optic system.

24
00:01:21,640 --> 00:01:25,080
And what I would like you to
do is to read that article

25
00:01:25,080 --> 00:01:28,650
but then proceed
through the internet

26
00:01:28,650 --> 00:01:34,330
to see what have become
the most recent discoveries

27
00:01:34,330 --> 00:01:37,760
about the so-called
accessory optic system.

28
00:01:37,760 --> 00:01:41,970
And that's what your
report needs to be about.

29
00:01:41,970 --> 00:01:48,110
You can be reasonably
summary about it

30
00:01:48,110 --> 00:01:50,352
so that you don't need
to write 50 pages.

31
00:01:50,352 --> 00:01:52,310
But you know, you can
write this whole thing up

32
00:01:52,310 --> 00:01:56,670
in four or five pages,
taking into account

33
00:01:56,670 --> 00:01:59,800
as I've said what
has been discovered

34
00:01:59,800 --> 00:02:03,600
since those days about the
so-called accessory optic

35
00:02:03,600 --> 00:02:04,590
system.

36
00:02:04,590 --> 00:02:07,920
The reason I thought this would
be a good assignment because it

37
00:02:07,920 --> 00:02:13,290
will give you a good historical
sense of how knowledge has

38
00:02:13,290 --> 00:02:19,100
expanded, in this
case, since the 1960s,

39
00:02:19,100 --> 00:02:23,410
in uncovering yet
another attribute

40
00:02:23,410 --> 00:02:26,170
that you have in the
visual system that

41
00:02:26,170 --> 00:02:30,640
starts with a specialized
set of cells in the retina

42
00:02:30,640 --> 00:02:37,020
and then proceeds through
a special series of steps

43
00:02:37,020 --> 00:02:43,860
until it connects with
the ocular motor system.

44
00:02:43,860 --> 00:02:45,320
So anyway, that's the task.

45
00:02:45,320 --> 00:02:51,900
And that will consist
of 10% of the report,

46
00:02:51,900 --> 00:02:53,630
of your total grade.

47
00:02:53,630 --> 00:02:57,660
And the same thing is
10% also for addition.

48
00:02:57,660 --> 00:03:01,630
Then the midterm exam
is going to consist

49
00:03:01,630 --> 00:03:08,060
of 25% of your exams.

50
00:03:08,060 --> 00:03:13,100
And the final exam is a total
55, a which 15 is vision

51
00:03:13,100 --> 00:03:14,760
and 40 is audition.

52
00:03:14,760 --> 00:03:17,590
So if you add them all up,
you have an equal distribution

53
00:03:17,590 --> 00:03:22,070
in percentages for
vision and for audition.

54
00:03:22,070 --> 00:03:25,180
So that then is
the basic layout.

55
00:03:25,180 --> 00:03:30,050
And so what I would like
you to start thinking about

56
00:03:30,050 --> 00:03:34,350
is how you're going to
put this report together.

57
00:03:34,350 --> 00:03:39,200
I was hoping that perhaps you
could get it done by midterm.

58
00:03:39,200 --> 00:03:42,670
But I don't have a hard
and fast rule about that.

59
00:03:42,670 --> 00:03:47,050
As long as you get it in before
the final exam, that's fine.

60
00:03:47,050 --> 00:03:49,350
But it may make
life easier if you

61
00:03:49,350 --> 00:03:53,740
can start working on that while
we are talking about vision,

62
00:03:53,740 --> 00:03:58,220
rather than waiting
until we are covering

63
00:03:58,220 --> 00:03:59,930
the auditory system
in the course.

64
00:03:59,930 --> 00:04:00,940

65
00:04:00,940 --> 00:04:04,655
Does anybody have any questions
about this basic layout?

66
00:04:04,655 --> 00:04:05,155
OK.

67
00:04:05,155 --> 00:04:25,660

68
00:04:25,660 --> 00:04:27,330
All right, now
today we are going

69
00:04:27,330 --> 00:04:29,520
to talk about form perception.

70
00:04:29,520 --> 00:04:33,950
Let me say one more thing,
that next time as you have it

71
00:04:33,950 --> 00:04:38,950
on the syllabus, we are going
to be covering illusions.

72
00:04:38,950 --> 00:04:39,970
I think that'll be fun.

73
00:04:39,970 --> 00:04:42,110
You'll see all kinds of
interesting illusions

74
00:04:42,110 --> 00:04:44,750
and some inferences
as to what we

75
00:04:44,750 --> 00:04:49,640
think about how those
illusions come about by virtue

76
00:04:49,640 --> 00:04:53,220
of the fact that there are
all sorts of interesting rules

77
00:04:53,220 --> 00:04:56,220
and facts about how the
visual system works.

78
00:04:56,220 --> 00:04:59,470
And then during the second
half over the next lecture,

79
00:04:59,470 --> 00:05:01,375
this is this coming
Wednesday, we

80
00:05:01,375 --> 00:05:03,466
are going to talk about
visual prosthesis.

81
00:05:03,466 --> 00:05:05,100

82
00:05:05,100 --> 00:05:07,090
All right, so that
in essence what is we

83
00:05:07,090 --> 00:05:08,410
are going to cover next time.

84
00:05:08,410 --> 00:05:10,259
So now let's get back to today.

85
00:05:10,259 --> 00:05:12,050
We are going to talk
about form perception.

86
00:05:12,050 --> 00:05:13,760

87
00:05:13,760 --> 00:05:18,870
This is a topic that,
even at this stage,

88
00:05:18,870 --> 00:05:21,670
we still only have
rather limited knowledge

89
00:05:21,670 --> 00:05:27,270
about how the brain is
capable of carrying out

90
00:05:27,270 --> 00:05:29,230
the perception of form.

91
00:05:29,230 --> 00:05:32,000
And we are going to look at
some inferences and some ideas

92
00:05:32,000 --> 00:05:32,860
about it.

93
00:05:32,860 --> 00:05:36,830
And I will also provide you with
a brief historical background.

94
00:05:36,830 --> 00:05:40,440
Because this has been a
topic of tremendous interest

95
00:05:40,440 --> 00:05:46,090
for people for centuries,
trying to understand just

96
00:05:46,090 --> 00:05:52,710
how we are capable of seeing
forms, shapes, patterns.

97
00:05:52,710 --> 00:05:57,810
All right, so the
first influential idea

98
00:05:57,810 --> 00:06:01,430
that emerged about this
is called structuralism.

99
00:06:01,430 --> 00:06:02,490

100
00:06:02,490 --> 00:06:09,210
And this idea actually took
place in the late 19th century.

101
00:06:09,210 --> 00:06:10,260

102
00:06:10,260 --> 00:06:14,890
And one of the prime proponents
was a fellow called Titchener.

103
00:06:14,890 --> 00:06:19,540
And they had, in trying to come
to grips with how we do this,

104
00:06:19,540 --> 00:06:24,150
had a rather basic
idea, much like building

105
00:06:24,150 --> 00:06:25,840
a house with bricks.

106
00:06:25,840 --> 00:06:29,970
The idea was that
perception is an aggregate

107
00:06:29,970 --> 00:06:31,049
of simple elements.

108
00:06:31,049 --> 00:06:33,590
We just have a whole bunch of
elements that you put together.

109
00:06:33,590 --> 00:06:38,310
And that generates a
more complex precept, OK.

110
00:06:38,310 --> 00:06:41,570
The problem with this approach,
well, the first problem

111
00:06:41,570 --> 00:06:44,306
was that they began
to do experiments.

112
00:06:44,306 --> 00:06:45,330

113
00:06:45,330 --> 00:06:49,640
And they asked subjects
to try to divorce

114
00:06:49,640 --> 00:06:52,140
a personal impression of
anything they looked at.

115
00:06:52,140 --> 00:06:54,400
But to sort of
physically describe it.

116
00:06:54,400 --> 00:06:56,990
So for example, if you
looked at an apple,

117
00:06:56,990 --> 00:07:01,350
how do you have the
idea that it's an apple?

118
00:07:01,350 --> 00:07:04,880
And so they said, well,
it has say, four colors.

119
00:07:04,880 --> 00:07:07,470
And so the person would
list four colors and said

120
00:07:07,470 --> 00:07:09,050
would have several shades.

121
00:07:09,050 --> 00:07:11,980
And put all those
together, add them all up,

122
00:07:11,980 --> 00:07:13,620
and that equals apple.

123
00:07:13,620 --> 00:07:14,700

124
00:07:14,700 --> 00:07:17,300
Well, that was the idea then.

125
00:07:17,300 --> 00:07:19,790
And when they did
this systematically,

126
00:07:19,790 --> 00:07:21,630
and the thing became
almost ridiculous.

127
00:07:21,630 --> 00:07:24,970
Because they came up
with more than 40,000

128
00:07:24,970 --> 00:07:28,790
elementary sensations,
thinking that somehow you

129
00:07:28,790 --> 00:07:31,800
puts these elementary
sensations together,

130
00:07:31,800 --> 00:07:33,910
40,000 of them,
and that gives us

131
00:07:33,910 --> 00:07:36,550
a sense of an apple,
or banana, or whatever.

132
00:07:36,550 --> 00:07:37,670

133
00:07:37,670 --> 00:07:42,260
Now as soon as this sort
of became well known,

134
00:07:42,260 --> 00:07:46,180
a lot of opposition arose.

135
00:07:46,180 --> 00:07:48,450
And one of the most
famous ones then

136
00:07:48,450 --> 00:07:50,360
was the opposition
that was brought

137
00:07:50,360 --> 00:07:51,973
about by Gestalt psychologists.

138
00:07:51,973 --> 00:07:52,980

139
00:07:52,980 --> 00:07:57,720
And the first consideration that
I want to sort of bring up here

140
00:07:57,720 --> 00:07:59,480
is a rather simple picture.

141
00:07:59,480 --> 00:08:01,800
Here's a picture
everybody looks at that.

142
00:08:01,800 --> 00:08:03,050
OK, what did you see there?

143
00:08:03,050 --> 00:08:04,780

144
00:08:04,780 --> 00:08:06,800
You saw a car, right?

145
00:08:06,800 --> 00:08:09,710
Now if you look at it,
I mean this is actually

146
00:08:09,710 --> 00:08:11,650
a picture of a car,
all right, but it's

147
00:08:11,650 --> 00:08:15,700
interrupted by all these
vertical bars here.

148
00:08:15,700 --> 00:08:20,320
And yet we're able to
infer that it's a car.

149
00:08:20,320 --> 00:08:24,090
And so the question comes
up, how do you do this?

150
00:08:24,090 --> 00:08:27,000
I mean, you can't simply
just add up a bunch of bricks

151
00:08:27,000 --> 00:08:28,310
and say it's a car.

152
00:08:28,310 --> 00:08:30,710
Because there's all
this interruption.

153
00:08:30,710 --> 00:08:34,190
And most things that
we see in the world,

154
00:08:34,190 --> 00:08:37,490
we see in a
discontinuous fashion.

155
00:08:37,490 --> 00:08:41,620
And as a result of this,
the Gestalt psychologists,

156
00:08:41,620 --> 00:08:45,830
which happened sort
of in the 1920s,

157
00:08:45,830 --> 00:08:47,860
began to think about
this whole problem

158
00:08:47,860 --> 00:08:49,470
in a very different way.

159
00:08:49,470 --> 00:08:51,145
And they made a
tremendous influence.

160
00:08:51,145 --> 00:08:52,530

161
00:08:52,530 --> 00:08:56,100
And they came up
with these new ideas

162
00:08:56,100 --> 00:08:59,830
of how we create the
perception of form.

163
00:08:59,830 --> 00:09:02,070
The founder of the
Gestalt psychology

164
00:09:02,070 --> 00:09:04,750
is a fellow called
Max Wertheimer.

165
00:09:04,750 --> 00:09:06,070

166
00:09:06,070 --> 00:09:10,270
So now, you can look this
up in Wikipedia, by the way,

167
00:09:10,270 --> 00:09:13,370
if you look at
them more closely,

168
00:09:13,370 --> 00:09:20,100
they came up with a few basic
principles of organization.

169
00:09:20,100 --> 00:09:22,930
And they call this grouping.

170
00:09:22,930 --> 00:09:26,600
They argue that in the brain,
somehow we group things up.

171
00:09:26,600 --> 00:09:29,800
And one of the things we group
has to do with proximity.

172
00:09:29,800 --> 00:09:32,500
Another one has to
do with similarity,

173
00:09:32,500 --> 00:09:36,500
another one with common motion,
another one with closure,

174
00:09:36,500 --> 00:09:39,050
and another one with
figure-ground perception.

175
00:09:39,050 --> 00:09:42,190
This last one I'm going to talk
about at the end of today's

176
00:09:42,190 --> 00:09:43,310
presentation.

177
00:09:43,310 --> 00:09:46,170
But now I'm going to give you
an example of these top two.

178
00:09:46,170 --> 00:09:48,560

179
00:09:48,560 --> 00:09:52,420
All right, so the most
important conclusion

180
00:09:52,420 --> 00:09:55,810
that they had come to which was
very much against structuralism

181
00:09:55,810 --> 00:09:58,680
is that the whole
that you see is

182
00:09:58,680 --> 00:10:00,880
different from the
sum of its parts.

183
00:10:00,880 --> 00:10:03,310
Somehow, there is
an active process

184
00:10:03,310 --> 00:10:07,490
that creates our ability to
see something that is not

185
00:10:07,490 --> 00:10:10,220
evidenced from its
individual parts.

186
00:10:10,220 --> 00:10:13,930
Now here is an example of
what these principles are.

187
00:10:13,930 --> 00:10:18,240
Grouping by proximity, so
here we have a bunch of dots.

188
00:10:18,240 --> 00:10:21,380
And if you'll put them
closer together vertically,

189
00:10:21,380 --> 00:10:24,150
you see a bunch of
vertical lines essentially.

190
00:10:24,150 --> 00:10:26,760
And if you put them closer
together horizontally,

191
00:10:26,760 --> 00:10:28,760
you see a bunch
of vertical lines,

192
00:10:28,760 --> 00:10:30,756
horizontal lines,
vertical lines.

193
00:10:30,756 --> 00:10:31,490
Yeah?

194
00:10:31,490 --> 00:10:36,980
So we group things
due to proximity.

195
00:10:36,980 --> 00:10:40,170
Another reason we group
things is shown here,

196
00:10:40,170 --> 00:10:45,610
is we group things according to
shape, or similarity of shape,

197
00:10:45,610 --> 00:10:47,000
I should really say.

198
00:10:47,000 --> 00:10:48,850
And here what we
have, you can readily

199
00:10:48,850 --> 00:10:58,190
see a group of nine disks here,
and a group of nine triangles.

200
00:10:58,190 --> 00:11:01,640
And if they're not all
the same as down here,

201
00:11:01,640 --> 00:11:04,460
then you have much more
difficulty grouping it.

202
00:11:04,460 --> 00:11:07,970
So there's a strong tendency
that we group things together

203
00:11:07,970 --> 00:11:09,440
that are similar.

204
00:11:09,440 --> 00:11:11,580
So these are some
general principles

205
00:11:11,580 --> 00:11:15,120
of how we organize
our visual percepts.

206
00:11:15,120 --> 00:11:17,680
So now we can look at some
more of these examples.

207
00:11:17,680 --> 00:11:20,830

208
00:11:20,830 --> 00:11:23,780
In doing this, we
are now going to try

209
00:11:23,780 --> 00:11:27,980
to be able to say a bit
more about the brain itself.

210
00:11:27,980 --> 00:11:33,030
And in doing so, it's evident
that the three major theories

211
00:11:33,030 --> 00:11:37,230
that try to deal with how
the brain does it, according

212
00:11:37,230 --> 00:11:40,070
to the first of
these, form perception

213
00:11:40,070 --> 00:11:43,610
is accomplished by neurons
that respond selectively

214
00:11:43,610 --> 00:11:46,280
to line segments of
different orientations.

215
00:11:46,280 --> 00:11:49,400
Now that theory,
obviously, is the outgrowth

216
00:11:49,400 --> 00:11:56,770
of what we talked about when it
was discovered that in V1, we

217
00:11:56,770 --> 00:11:59,750
have orientation
selected neurons.

218
00:11:59,750 --> 00:12:01,930
And this orientation
selectivity is something

219
00:12:01,930 --> 00:12:04,230
that you see in
several progressively

220
00:12:04,230 --> 00:12:08,400
higher visual areas, as if
everything in the world out

221
00:12:08,400 --> 00:12:12,190
there was broken down into
oriented line segments, which

222
00:12:12,190 --> 00:12:16,750
are then somehow put together
according to there orientation

223
00:12:16,750 --> 00:12:18,210
to enable us to see shapes.

224
00:12:18,210 --> 00:12:19,620
So that's one theory.

225
00:12:19,620 --> 00:12:23,500
Another theory is that form
perception is accomplished

226
00:12:23,500 --> 00:12:29,490
by spatial mapping of the visual
scene onto the visual cortex.

227
00:12:29,490 --> 00:12:31,580
And I will elaborate
on each of these.

228
00:12:31,580 --> 00:12:34,960
And the third theory is one that
form perception is accomplished

229
00:12:34,960 --> 00:12:38,710
by virtue of Fourier analysis.

230
00:12:38,710 --> 00:12:44,580
So let us first then
turn to form perception,

231
00:12:44,580 --> 00:12:49,750
supposedly based on breaking
down the visual scene

232
00:12:49,750 --> 00:12:52,180
into oriented line segments.

233
00:12:52,180 --> 00:12:54,940
Now, one of the big
problems with this

234
00:12:54,940 --> 00:13:03,300
is that when you look
at a famous art work

235
00:13:03,300 --> 00:13:10,920
that you see in every day
in the Wall Street Journal,

236
00:13:10,920 --> 00:13:13,370
there is an artist who
created this originally.

237
00:13:13,370 --> 00:13:17,750
And he created faces
like this using only dots

238
00:13:17,750 --> 00:13:20,120
and varied the spatial
frequency of the dots,

239
00:13:20,120 --> 00:13:21,360
as you can see here.

240
00:13:21,360 --> 00:13:22,410

241
00:13:22,410 --> 00:13:25,150
This happens to be a
person called Larry Poons.

242
00:13:25,150 --> 00:13:27,120
I'm sure none of you
have heard of him.

243
00:13:27,120 --> 00:13:30,680
But here was one of the
pictures many years ago

244
00:13:30,680 --> 00:13:32,280
in the Wall Street Journal.

245
00:13:32,280 --> 00:13:36,150
And even today, as I say, you
can see faces created this way.

246
00:13:36,150 --> 00:13:38,700
And you can readily
recognize them,

247
00:13:38,700 --> 00:13:41,960
even though there are no
oriented line segments here.

248
00:13:41,960 --> 00:13:46,300
And in fact, if you sort
of squint up and you three

249
00:13:46,300 --> 00:13:48,320
quarter ways close your
eyes, so you can't even

250
00:13:48,320 --> 00:13:50,720
see the dots anymore,
you can still

251
00:13:50,720 --> 00:13:52,670
make out that face
very, very clearly.

252
00:13:52,670 --> 00:13:54,000

253
00:13:54,000 --> 00:13:57,330
That's because the spatial
frequency effect here

254
00:13:57,330 --> 00:13:58,880
is very important,
and the degree

255
00:13:58,880 --> 00:14:02,780
of shading, meaning that
this involves rather low

256
00:14:02,780 --> 00:14:05,650
spatial frequency
analyses, rather than

257
00:14:05,650 --> 00:14:11,870
the analysis of particular
orientations of line segments.

258
00:14:11,870 --> 00:14:21,080
Now this particular
analysis has its counterpart

259
00:14:21,080 --> 00:14:23,900
in the observation that
many of you have seen.

260
00:14:23,900 --> 00:14:29,670
When you look at a person's
face on television,

261
00:14:29,670 --> 00:14:31,750
and they want to
prevent you from being

262
00:14:31,750 --> 00:14:35,184
able to see that face because
it's confidential or something,

263
00:14:35,184 --> 00:14:35,850
what do they do?

264
00:14:35,850 --> 00:14:36,987

265
00:14:36,987 --> 00:14:37,695
Anybody remember?

266
00:14:37,695 --> 00:14:39,630

267
00:14:39,630 --> 00:14:41,240
What you do is
you put up a bunch

268
00:14:41,240 --> 00:14:47,890
of squares of different
spatial frequencies.

269
00:14:47,890 --> 00:14:52,560
And each square is comprised
of the mean illumination

270
00:14:52,560 --> 00:14:55,170
level of the actual face.

271
00:14:55,170 --> 00:14:58,930
If you do that, this high
frequency information

272
00:14:58,930 --> 00:15:01,890
is something that
interferes with your ability

273
00:15:01,890 --> 00:15:03,210
to analyze face.

274
00:15:03,210 --> 00:15:05,850
So to give you an example
of this, I'm sure all of you

275
00:15:05,850 --> 00:15:08,590
have seen this, but
mainly I show you this.

276
00:15:08,590 --> 00:15:10,280
Here we have this example.

277
00:15:10,280 --> 00:15:14,010
How many of you can tell
who these people are here?

278
00:15:14,010 --> 00:15:15,050

279
00:15:15,050 --> 00:15:17,350
I guess nobody can, right?

280
00:15:17,350 --> 00:15:20,780
Now then what we
can do is we can

281
00:15:20,780 --> 00:15:23,250
increase the frequency of these.

282
00:15:23,250 --> 00:15:24,250
Here it is.

283
00:15:24,250 --> 00:15:26,550
And it's still very
difficult to tell.

284
00:15:26,550 --> 00:15:29,180
And now I'm going to show
you the actual photograph.

285
00:15:29,180 --> 00:15:32,710
And what I'm curious about is
how many of these two people

286
00:15:32,710 --> 00:15:33,700
you actually know.

287
00:15:33,700 --> 00:15:34,720

288
00:15:34,720 --> 00:15:36,200
I can tell you
right away, they're

289
00:15:36,200 --> 00:15:41,520
extremely famous
actors and actresses.

290
00:15:41,520 --> 00:15:44,217
Who recognizes these people?

291
00:15:44,217 --> 00:15:45,300
How many of you recognize?

292
00:15:45,300 --> 00:15:46,630
Let me see your hand.

293
00:15:46,630 --> 00:15:48,480
Only a few people recognize it.

294
00:15:48,480 --> 00:15:51,330
That's how quickly time goes by.

295
00:15:51,330 --> 00:15:54,600
These were the most central,
most exciting, best known

296
00:15:54,600 --> 00:15:57,040
people in the movies.

297
00:15:57,040 --> 00:15:58,120

298
00:15:58,120 --> 00:16:04,560
And this is Humphrey
Bogart, Ingrid Bergman.

299
00:16:04,560 --> 00:16:07,760
OK, so these are these
two very famous people

300
00:16:07,760 --> 00:16:12,110
whose faces were obstructed
by this and by this using

301
00:16:12,110 --> 00:16:18,550
these high contrast edges
that obscure your ability

302
00:16:18,550 --> 00:16:21,980
to smoothly analyze faces.

303
00:16:21,980 --> 00:16:33,950
So that then suggests that this
idea that we extensively use

304
00:16:33,950 --> 00:16:43,790
oriented line segments
to analyze faces at least

305
00:16:43,790 --> 00:16:47,570
is an insufficient
explanation of how

306
00:16:47,570 --> 00:16:51,790
the brain processes shapes.

307
00:16:51,790 --> 00:16:55,565
Now another idea is so-called
topographic mapping.

308
00:16:55,565 --> 00:16:57,440

309
00:16:57,440 --> 00:17:00,080
This idea I'm sure
you can already

310
00:17:00,080 --> 00:17:02,480
reject on the basis
of what I told you

311
00:17:02,480 --> 00:17:06,960
about how the visual
system is laid out.

312
00:17:06,960 --> 00:17:08,720
But now I'm going
to belabor that

313
00:17:08,720 --> 00:17:10,650
so that you can
follow it closely.

314
00:17:10,650 --> 00:17:13,369
Here we have a
runner, of course.

315
00:17:13,369 --> 00:17:17,079
And here we have a monkey brain
to make it easy to understand.

316
00:17:17,079 --> 00:17:21,140
And the idea here there is
that this image, once it

317
00:17:21,140 --> 00:17:25,770
was discovered that the
visual field is laid out

318
00:17:25,770 --> 00:17:29,030
topographically in
the visual cortex,

319
00:17:29,030 --> 00:17:33,800
the idea was that what
happens is somehow,

320
00:17:33,800 --> 00:17:37,480
and I may be am unfair
to poke fun of it,

321
00:17:37,480 --> 00:17:41,240
but the idea is
that the mind can

322
00:17:41,240 --> 00:17:47,150
look at the creation of this
image on the cortical surface

323
00:17:47,150 --> 00:17:47,650
there.

324
00:17:47,650 --> 00:17:49,130

325
00:17:49,130 --> 00:17:52,190
And thereby, it can identify.

326
00:17:52,190 --> 00:17:55,180
OK, it's almost
like a photograph,

327
00:17:55,180 --> 00:17:56,690
looking at a photograph.

328
00:17:56,690 --> 00:18:01,320
In this case the mind looks at
the photograph, so to speak,

329
00:18:01,320 --> 00:18:02,850
on the cortical surface.

330
00:18:02,850 --> 00:18:05,470
So they thought at
the time is that this

331
00:18:05,470 --> 00:18:07,250
is what you have there, OK.

332
00:18:07,250 --> 00:18:10,310
That's the imprint
of this image.

333
00:18:10,310 --> 00:18:12,290
And indeed, you say,
oh my goodness, that's

334
00:18:12,290 --> 00:18:13,000
just like that.

335
00:18:13,000 --> 00:18:17,210
Therefore, I can recognize that
person, and so on and so on.

336
00:18:17,210 --> 00:18:20,120
Well, I mean that's a cute idea.

337
00:18:20,120 --> 00:18:22,400
But then when you
take into account

338
00:18:22,400 --> 00:18:27,280
this has been discovered
subsequently to these ideas,

339
00:18:27,280 --> 00:18:33,700
that the topographic
layout in the visual cortex

340
00:18:33,700 --> 00:18:41,320
is actually not one to one
because of the magnification

341
00:18:41,320 --> 00:18:42,450
factor.

342
00:18:42,450 --> 00:18:45,060
So let's look at
that in some detail.

343
00:18:45,060 --> 00:18:51,410
Here is an actual reconstruction
of the monkey area V1 here.

344
00:18:51,410 --> 00:18:52,490

345
00:18:52,490 --> 00:18:55,170
And here is the visual field.

346
00:18:55,170 --> 00:18:58,665
And we are going to
put these red arrows

347
00:18:58,665 --> 00:19:00,740
in the contralateral hemifield.

348
00:19:00,740 --> 00:19:03,030
Now remember what
I told you before,

349
00:19:03,030 --> 00:19:06,120
that if you look at
the visual scene,

350
00:19:06,120 --> 00:19:09,210
you can imagine your eyes
being vertically cut in half.

351
00:19:09,210 --> 00:19:12,350
And you have a nasal
and temporal hemiretina.

352
00:19:12,350 --> 00:19:18,360
And the one which is
contralateral hemifield

353
00:19:18,360 --> 00:19:22,270
crosses over and gets into
this half of the brain.

354
00:19:22,270 --> 00:19:25,100
And this one crosses
over to that half.

355
00:19:25,100 --> 00:19:28,720
So now if you do this and
take the magnification factor

356
00:19:28,720 --> 00:19:33,750
into account, you
look at these arrows,

357
00:19:33,750 --> 00:19:35,850
which are all
identical in size, OK.

358
00:19:35,850 --> 00:19:39,430
There are one, two, three,
four, five, six, seven of them.

359
00:19:39,430 --> 00:19:43,940
What you can see here
is the actual impression

360
00:19:43,940 --> 00:19:46,310
on the cortical
surface of the neurons

361
00:19:46,310 --> 00:19:50,250
that are being activated
by this look like this,

362
00:19:50,250 --> 00:19:53,480
nothing like those arrows there.

363
00:19:53,480 --> 00:19:56,300
And in fact, the central
arrow is much bigger

364
00:19:56,300 --> 00:19:59,420
because of the huge
magnification factor.

365
00:19:59,420 --> 00:20:05,790
Now this is already
creates a major problem

366
00:20:05,790 --> 00:20:11,890
in trying to believe in
that particular theory,

367
00:20:11,890 --> 00:20:13,910
the topographic mapping theory.

368
00:20:13,910 --> 00:20:18,180
But now, if instead,
you put those arrows

369
00:20:18,180 --> 00:20:21,520
halfway across each of
the hemispheres like that,

370
00:20:21,520 --> 00:20:23,720
then whatever is on
this side goes here.

371
00:20:23,720 --> 00:20:25,570
Whatever is on this
side goes there.

372
00:20:25,570 --> 00:20:28,140
And this is the impression
that is created.

373
00:20:28,140 --> 00:20:29,600

374
00:20:29,600 --> 00:20:31,996
And my goodness, you
create that impression.

375
00:20:31,996 --> 00:20:34,030

376
00:20:34,030 --> 00:20:36,234
And you say to yourself, my god.

377
00:20:36,234 --> 00:20:37,650
If that's the case,
how come I can

378
00:20:37,650 --> 00:20:41,240
see seven straight
arrows of equal size

379
00:20:41,240 --> 00:20:44,740
when this is the impression that
is being created in the brain?

380
00:20:44,740 --> 00:20:52,080
So therefore, obviously, they
are not using a topographic map

381
00:20:52,080 --> 00:20:54,180
to analyze the
visual scene at all.

382
00:20:54,180 --> 00:20:56,150

383
00:20:56,150 --> 00:20:58,800
Now this can be driven
home even further.

384
00:20:58,800 --> 00:21:00,780
Let me make another point here.

385
00:21:00,780 --> 00:21:04,670
Here we have a monkey visual
cortex again from the rear.

386
00:21:04,670 --> 00:21:06,600
Here is the visual
field laid out.

387
00:21:06,600 --> 00:21:11,890
And if here we put in a bunch
of dots in a circular fashion,

388
00:21:11,890 --> 00:21:14,037
this is the activation there.

389
00:21:14,037 --> 00:21:15,370
Now you say, oh, that's not bad.

390
00:21:15,370 --> 00:21:17,260
That looks like a circle.

391
00:21:17,260 --> 00:21:23,060
But then if you put it along
the midline, half and half,

392
00:21:23,060 --> 00:21:25,300
then this is the
actual activation.

393
00:21:25,300 --> 00:21:29,100
But you still see a circle,
even though the activation,

394
00:21:29,100 --> 00:21:32,910
in terms of the topography,
is nothing like a circle.

395
00:21:32,910 --> 00:21:40,300
So that indeed created a
much, much greater degree

396
00:21:40,300 --> 00:21:43,270
of skepticism
among investigators

397
00:21:43,270 --> 00:21:46,220
to try to understand
how we process shape.

398
00:21:46,220 --> 00:21:47,546

399
00:21:47,546 --> 00:21:49,790
Now let me make one
other point here,

400
00:21:49,790 --> 00:21:52,060
which is a wonderful
story, which

401
00:21:52,060 --> 00:21:55,320
I think will be good
for you to remember.

402
00:21:55,320 --> 00:21:57,530
That's called the
Giotto story, which

403
00:21:57,530 --> 00:21:59,850
says that when Pope
Benedict, that was

404
00:21:59,850 --> 00:22:05,610
in the 12th century,
or the 13th century,

405
00:22:05,610 --> 00:22:09,620
he set out to have the walls
of the great cathedral of Saint

406
00:22:09,620 --> 00:22:13,740
Peter in Rome redecorated.

407
00:22:13,740 --> 00:22:17,510
And so he sent out a
bunch of messengers

408
00:22:17,510 --> 00:22:21,570
to various artists
in Italy and asked

409
00:22:21,570 --> 00:22:24,560
them to provide some
of their best work

410
00:22:24,560 --> 00:22:26,530
so he could evaluate
it and could

411
00:22:26,530 --> 00:22:30,005
pick one guy to actually
do the redecoration.

412
00:22:30,005 --> 00:22:31,010

413
00:22:31,010 --> 00:22:37,520
Well, one of these messengers
went to Ambrogio Bondone Giotto

414
00:22:37,520 --> 00:22:40,290
and asked him, he was
a well known artist,

415
00:22:40,290 --> 00:22:44,830
and asked him to provide
a painting of his,

416
00:22:44,830 --> 00:22:46,270
a drawing of his.

417
00:22:46,270 --> 00:22:49,210
And Giotto said,
oh, my goodness.

418
00:22:49,210 --> 00:22:50,760
I just don't have
anything around.

419
00:22:50,760 --> 00:22:52,420
But I tell you what I'll do.

420
00:22:52,420 --> 00:22:57,180
He took out a red pen and
drew a perfect circle, OK.

421
00:22:57,180 --> 00:23:01,370
And so the messenger
took this to the pope.

422
00:23:01,370 --> 00:23:02,700
And the pope said, my god.

423
00:23:02,700 --> 00:23:04,700
I can't believe how
incredible this is.

424
00:23:04,700 --> 00:23:09,610
And so Giotto got the
job of redecorating

425
00:23:09,610 --> 00:23:13,630
the cathedral of Saint Peter.

426
00:23:13,630 --> 00:23:18,820
To this day, there is
this expression in Italy,

427
00:23:18,820 --> 00:23:25,390
in Tuscany in Italy, which
says the round O of Giotto, OK.

428
00:23:25,390 --> 00:23:27,810
So this is the
round O of Giotto,

429
00:23:27,810 --> 00:23:32,850
which somehow, in a sense,
denotes perfection, perfection

430
00:23:32,850 --> 00:23:38,580
in sight and in perfection
in your ability of execution

431
00:23:38,580 --> 00:23:40,890
in terms of a
drawing, for example.

432
00:23:40,890 --> 00:23:44,350
All right, now to highlight
this even further,

433
00:23:44,350 --> 00:23:46,470
let me point out
this to you here.

434
00:23:46,470 --> 00:23:53,040
What we have here is a bunch of
imperfect circles, one of which

435
00:23:53,040 --> 00:23:54,810
is perfect.

436
00:23:54,810 --> 00:23:56,810
So if you keep
looking around, you

437
00:23:56,810 --> 00:23:58,920
should be able to
spot which one it is.

438
00:23:58,920 --> 00:24:02,170
And you should tell me what
letter denotes that circle.

439
00:24:02,170 --> 00:24:03,420
Which one is a perfect circle?

440
00:24:03,420 --> 00:24:06,000

441
00:24:06,000 --> 00:24:06,790
AUDIENCE: C.

442
00:24:06,790 --> 00:24:08,510
PROFESSOR: Very good, all right.

443
00:24:08,510 --> 00:24:10,370
So we are incredibly good.

444
00:24:10,370 --> 00:24:13,320
There's a slight difference
here, OK, or even slight

445
00:24:13,320 --> 00:24:14,280
difference here.

446
00:24:14,280 --> 00:24:17,060
And yet, we can see this
very slight difference.

447
00:24:17,060 --> 00:24:18,330

448
00:24:18,330 --> 00:24:20,520
And we can tell what
a perfect circle is.

449
00:24:20,520 --> 00:24:24,890
I mean, that's incredible
given how the impressions are

450
00:24:24,890 --> 00:24:29,380
made by those circles
on the visual cortex.

451
00:24:29,380 --> 00:24:31,685
So that is an
incredible puzzle of how

452
00:24:31,685 --> 00:24:33,570
we are capable of doing this.

453
00:24:33,570 --> 00:24:35,370
And I'm afraid,
even to this day,

454
00:24:35,370 --> 00:24:39,110
we don't have a really good
answer of how this happens.

455
00:24:39,110 --> 00:24:50,340
OK, now the third theory
that has become actually

456
00:24:50,340 --> 00:24:53,940
probably right now, one of
the most successful ones,

457
00:24:53,940 --> 00:24:59,540
it claims that you
analyze the visual scene

458
00:24:59,540 --> 00:25:04,950
by taking into account the
spatial frequencies that

459
00:25:04,950 --> 00:25:08,590
are impinging on the retina, OK.

460
00:25:08,590 --> 00:25:11,470
This was created by Fergus
Campbell and John Robson,

461
00:25:11,470 --> 00:25:12,950
very influential.

462
00:25:12,950 --> 00:25:17,230
And they pointed out, first of
all, that a very interesting

463
00:25:17,230 --> 00:25:19,880
finding, which I think
I've mentioned once

464
00:25:19,880 --> 00:25:23,590
before, that if you vary the
spatial frequency [? both, ?]

465
00:25:23,590 --> 00:25:27,790
as well as the contrast, we
have this sensitivity function

466
00:25:27,790 --> 00:25:28,700
like this.

467
00:25:28,700 --> 00:25:30,990
I did show you a picture
of it the last time.

468
00:25:30,990 --> 00:25:32,370

469
00:25:32,370 --> 00:25:35,630
And then, furthermore,
what they had shown

470
00:25:35,630 --> 00:25:41,660
is that you can create all
kinds of complex precepts

471
00:25:41,660 --> 00:25:44,630
by varying the gratings.

472
00:25:44,630 --> 00:25:47,440
Make them simple gratings,
compound gratings,

473
00:25:47,440 --> 00:25:51,110
and compound gratings
with much lower contrast.

474
00:25:51,110 --> 00:25:54,194
And this is simple different
spatial frequencies.

475
00:25:54,194 --> 00:25:54,985
These are compound.

476
00:25:54,985 --> 00:25:56,140

477
00:25:56,140 --> 00:25:59,390
And these, as I've said, are
lower spatial frequencies.

478
00:25:59,390 --> 00:26:02,190
So then if you again
squint, then you

479
00:26:02,190 --> 00:26:03,610
will see something
much smoother.

480
00:26:03,610 --> 00:26:04,790

481
00:26:04,790 --> 00:26:09,260
So they actually carried out a
detailed mathematical analysis

482
00:26:09,260 --> 00:26:11,480
using Fourier analysis.

483
00:26:11,480 --> 00:26:14,300
And what they did
was quite remarkable.

484
00:26:14,300 --> 00:26:17,820
They would down break
down a visual scene,

485
00:26:17,820 --> 00:26:23,270
like a photograph of New
York with all the skylights,

486
00:26:23,270 --> 00:26:25,580
OK, skyscrapers.

487
00:26:25,580 --> 00:26:30,850
And then they would convert
that using Fourier analysis.

488
00:26:30,850 --> 00:26:33,860
And they could recreate
the visual scene

489
00:26:33,860 --> 00:26:36,955
with a high degree of
accuracy using that procedure.

490
00:26:36,955 --> 00:26:38,150

491
00:26:38,150 --> 00:26:42,810
Now if indeed the
visual system uses this,

492
00:26:42,810 --> 00:26:46,390
you have a number of basic
logical requirements.

493
00:26:46,390 --> 00:26:50,340
And those basic logical,
let me come back to that,

494
00:26:50,340 --> 00:26:52,890
the basic logical
requirements are

495
00:26:52,890 --> 00:26:57,220
that you need spatial
frequency analysis.

496
00:26:57,220 --> 00:27:00,770
And I've shown you that already
when we talked about V1,

497
00:27:00,770 --> 00:27:05,680
that neurons there are
spatial frequency selective.

498
00:27:05,680 --> 00:27:09,000
Secondly, that there
are contrast selective,

499
00:27:09,000 --> 00:27:10,300
which you know already.

500
00:27:10,300 --> 00:27:13,270
And of course, the
orientation selective,

501
00:27:13,270 --> 00:27:16,830
and they can tell
you about phase.

502
00:27:16,830 --> 00:27:19,410
As long as you have
these four attributes,

503
00:27:19,410 --> 00:27:22,840
you can perform a
detailed Fourier analysis

504
00:27:22,840 --> 00:27:25,170
to reconstruct the visual scene.

505
00:27:25,170 --> 00:27:28,270
So now to stress
this even further,

506
00:27:28,270 --> 00:27:32,270
they did a series of
experiments, in which they

507
00:27:32,270 --> 00:27:35,420
asked a question,
is it true that you

508
00:27:35,420 --> 00:27:37,710
have a particular spatial
frequency analysis,

509
00:27:37,710 --> 00:27:39,570
that you can manipulate that?

510
00:27:39,570 --> 00:27:42,490
And so they did an
experiment, which

511
00:27:42,490 --> 00:27:46,130
called a frequency-specific
adaptation experiment.

512
00:27:46,130 --> 00:27:49,530
They would present to
a subject this display

513
00:27:49,530 --> 00:27:52,840
and have the subject look at
this for a couple of minutes

514
00:27:52,840 --> 00:27:54,670
without having to fixate.

515
00:27:54,670 --> 00:27:56,760
And then they would
look at each of those.

516
00:27:56,760 --> 00:28:00,150
And they found that this, which
is the same spatial frequency,

517
00:28:00,150 --> 00:28:03,890
they had difficulty seeing
because of this adaptation.

518
00:28:03,890 --> 00:28:07,080
And so then they did a
series of careful studies

519
00:28:07,080 --> 00:28:09,420
and carried out this
analysis and showed

520
00:28:09,420 --> 00:28:13,250
that you could get any
kind of spatial frequency

521
00:28:13,250 --> 00:28:16,240
to lose your sensitivity
for it if you

522
00:28:16,240 --> 00:28:18,170
had been pre-exposed to it.

523
00:28:18,170 --> 00:28:19,860
So that's what that looked like.

524
00:28:19,860 --> 00:28:22,070
And by doing this
systematically,

525
00:28:22,070 --> 00:28:24,880
they came up with
the idea that what

526
00:28:24,880 --> 00:28:31,080
you have in the visual cortex
is a series of channels

527
00:28:31,080 --> 00:28:33,765
that are spatial
frequency-selective.

528
00:28:33,765 --> 00:28:34,850

529
00:28:34,850 --> 00:28:37,610
And I showed you
we talked about V1,

530
00:28:37,610 --> 00:28:41,760
that indeed, there are neurons
there that are selected

531
00:28:41,760 --> 00:28:43,640
to particular
spatial frequencies.

532
00:28:43,640 --> 00:28:47,510
And they proposed that you have
a series of channels like this,

533
00:28:47,510 --> 00:28:51,550
OK, that peak at different
spatial frequencies.

534
00:28:51,550 --> 00:28:52,740

535
00:28:52,740 --> 00:28:55,160
And by activating
them selectively,

536
00:28:55,160 --> 00:28:57,830
you can reconstruct
virtually anything out

537
00:28:57,830 --> 00:29:00,750
in the visual scene
using Fourier analysis.

538
00:29:00,750 --> 00:29:11,080
Now OK, I think that then is
the essence of that theory.

539
00:29:11,080 --> 00:29:14,550
And I can tell you that some
people avidly believe in it.

540
00:29:14,550 --> 00:29:17,790
And there are some people
who are highly skeptical.

541
00:29:17,790 --> 00:29:20,320
I'm not sure where I stand
at this stage about that.

542
00:29:20,320 --> 00:29:24,720

543
00:29:24,720 --> 00:29:30,330
But now, people began
to study our shape,

544
00:29:30,330 --> 00:29:36,480
the ability to see shapes by
recording various visual areas.

545
00:29:36,480 --> 00:29:40,450
And I want to tell you next
about some studies that

546
00:29:40,450 --> 00:29:43,450
had been done primarily
in Japan looking

547
00:29:43,450 --> 00:29:47,900
at inferotemporal cortex,
which is [INAUDIBLE] already

548
00:29:47,900 --> 00:29:52,680
mentioned that to be involved
in the analysis of shapes

549
00:29:52,680 --> 00:29:57,430
and in particular, in the
analysis of faces as well.

550
00:29:57,430 --> 00:30:01,150
So what these
investigators did is

551
00:30:01,150 --> 00:30:03,910
they would record to
individual neurons

552
00:30:03,910 --> 00:30:07,735
in inferotemporal cortex.

553
00:30:07,735 --> 00:30:09,030

554
00:30:09,030 --> 00:30:13,730
And when they did that, they
would present various stimuli

555
00:30:13,730 --> 00:30:15,135
to see how those
cells responded.

556
00:30:15,135 --> 00:30:16,590

557
00:30:16,590 --> 00:30:19,630
And they found, they
claimed to find,

558
00:30:19,630 --> 00:30:22,750
that there was some
incredible specificity

559
00:30:22,750 --> 00:30:26,370
in inferotemporal
cortex for shapes.

560
00:30:26,370 --> 00:30:29,120
And so what they did,
here's an example.

561
00:30:29,120 --> 00:30:32,430
Here, we have the neurons'
responses, histograms.

562
00:30:32,430 --> 00:30:35,040
And on top, above it, we
have the particular shape

563
00:30:35,040 --> 00:30:37,000
that was presented repeatedly.

564
00:30:37,000 --> 00:30:40,940
So this particular cell
responded vigorously to this,

565
00:30:40,940 --> 00:30:44,370
but very poorly to that,
and so on down the line.

566
00:30:44,370 --> 00:30:46,090
But you can see
that it responded

567
00:30:46,090 --> 00:30:48,370
to several different shapes,
quite well to this one,

568
00:30:48,370 --> 00:30:52,480
reasonably well to that one,
so a whole range of them.

569
00:30:52,480 --> 00:30:53,490

570
00:30:53,490 --> 00:30:57,650
So it wasn't like this
particular neuron responded

571
00:30:57,650 --> 00:31:00,340
only to one particular shape.

572
00:31:00,340 --> 00:31:02,670
But that idea that
you have neurons

573
00:31:02,670 --> 00:31:05,980
which are specific
for certain shapes

574
00:31:05,980 --> 00:31:12,990
did take on a lot of attraction
on part of investigators.

575
00:31:12,990 --> 00:31:15,540
And some of them
indeed thought that you

576
00:31:15,540 --> 00:31:20,180
have these neurons
which are selective

577
00:31:20,180 --> 00:31:23,050
to individual elements
in the visual scene.

578
00:31:23,050 --> 00:31:24,490

579
00:31:24,490 --> 00:31:29,790
And those subsequently
the critics refer to

580
00:31:29,790 --> 00:31:33,050
as your having
grandmother cells.

581
00:31:33,050 --> 00:31:35,100
Somewhere in the
brain there's a cell

582
00:31:35,100 --> 00:31:37,900
that represents your
grandmother, all right.

583
00:31:37,900 --> 00:31:41,530
Now that idea was
disabuse subsequently.

584
00:31:41,530 --> 00:31:46,620
But it took very stronghold
in many investigators' minds.

585
00:31:46,620 --> 00:31:48,600
And here's another
example of yet

586
00:31:48,600 --> 00:31:51,260
another inferotemporal
neurons, in this case again

587
00:31:51,260 --> 00:31:55,450
for various shapes, showing
that this shape was,

588
00:31:55,450 --> 00:31:58,000
elicited a lot of responses,
as well as did this.

589
00:31:58,000 --> 00:32:00,950
The rest of them didn't
respond, didn't elicit as much

590
00:32:00,950 --> 00:32:02,370
of a response.

591
00:32:02,370 --> 00:32:05,760
So these kinds of
experiments then,

592
00:32:05,760 --> 00:32:06,843
they tried to systematize.

593
00:32:06,843 --> 00:32:07,860

594
00:32:07,860 --> 00:32:12,810
And they came up with an idea
which, at least to my mind,

595
00:32:12,810 --> 00:32:16,070
borders on the absurd,
which is shown here.

596
00:32:16,070 --> 00:32:18,750
Here is a section of
inferotemporal cortex.

597
00:32:18,750 --> 00:32:21,380
And they argue that
they are columns there,

598
00:32:21,380 --> 00:32:25,610
and that these columns
represent different percepts.

599
00:32:25,610 --> 00:32:27,780
So here we have a
bunch of percepts

600
00:32:27,780 --> 00:32:28,970
that process the monkey.

601
00:32:28,970 --> 00:32:33,470
This is a monkey, of course,
that process monkey faces.

602
00:32:33,470 --> 00:32:35,070
And by inference,
therefore, there

603
00:32:35,070 --> 00:32:38,330
must be some areas in
inferotemporal cortex

604
00:32:38,330 --> 00:32:42,260
that process human
faces in humans.

605
00:32:42,260 --> 00:32:44,790
And then others process
different aspects

606
00:32:44,790 --> 00:32:45,770
of the visual scene.

607
00:32:45,770 --> 00:32:46,970

608
00:32:46,970 --> 00:32:51,430
Now the big problem, technical
problem, with this approach

609
00:32:51,430 --> 00:32:55,530
is that when you record
from individual neurons,

610
00:32:55,530 --> 00:32:57,780
all right, typically
in these experiments,

611
00:32:57,780 --> 00:33:01,790
you can study a single neuron
for a fairly limited time,

612
00:33:01,790 --> 00:33:06,890
not like for months on end, just
maybe a few hours or something.

613
00:33:06,890 --> 00:33:09,150
And so because of
that, you can only

614
00:33:09,150 --> 00:33:12,890
present a limited number
of visual stimuli.

615
00:33:12,890 --> 00:33:16,510
Now there are millions of
visual stimuli out there.

616
00:33:16,510 --> 00:33:19,850
And so to really establish how
specific this particular neuron

617
00:33:19,850 --> 00:33:22,920
is, they say well, these are
the only shapes it's showing us.

618
00:33:22,920 --> 00:33:28,004
What if you use the
cross, or who knows,

619
00:33:28,004 --> 00:33:29,420
many, many other
things, something

620
00:33:29,420 --> 00:33:30,990
that is three dimensional?

621
00:33:30,990 --> 00:33:32,910
How would these cells respond?

622
00:33:32,910 --> 00:33:36,297
And so these cells respond
to different degrees,

623
00:33:36,297 --> 00:33:38,755
to hundreds and hundreds and
hundreds of different stimuli.

624
00:33:38,755 --> 00:33:39,860

625
00:33:39,860 --> 00:33:44,440
And the real fact then is
that anytime a stimulus

626
00:33:44,440 --> 00:33:48,290
appears out there, you're
activating tens of thousands

627
00:33:48,290 --> 00:33:53,580
of neurons, maybe even
more, in the visual system.

628
00:33:53,580 --> 00:33:55,400
And each of those
neurons, especially

629
00:33:55,400 --> 00:34:00,450
in inferotemporal cortex,
responds to different degrees

630
00:34:00,450 --> 00:34:02,170
to the different stimuli.

631
00:34:02,170 --> 00:34:07,680
It is the compendium of these
many, many neurons firing

632
00:34:07,680 --> 00:34:10,590
into different degrees
to different stimuli

633
00:34:10,590 --> 00:34:14,400
that gives you sort of an
overall computational ability

634
00:34:14,400 --> 00:34:16,810
to say what that stimulus is.

635
00:34:16,810 --> 00:34:20,370
Now to be able to analyze
that, it is that complicated,

636
00:34:20,370 --> 00:34:22,989
that really takes
a lot of effort.

637
00:34:22,989 --> 00:34:25,630
Now some people are
now trying to do this.

638
00:34:25,630 --> 00:34:29,389
And the way you
try to do this is

639
00:34:29,389 --> 00:34:31,389
that you're recording
from hundreds and hundreds

640
00:34:31,389 --> 00:34:35,620
of neurons with
multiple electrodes,

641
00:34:35,620 --> 00:34:37,820
present various scenes.

642
00:34:37,820 --> 00:34:42,070
And you see how these neurons
respond as an aggregate,

643
00:34:42,070 --> 00:34:45,400
so that you can determine
whether or not there is indeed

644
00:34:45,400 --> 00:34:48,880
the potential for some
sort of computation

645
00:34:48,880 --> 00:34:52,929
that takes place in
gaining the impression

646
00:34:52,929 --> 00:34:56,570
of a particular individual face.

647
00:34:56,570 --> 00:34:59,340
Now one of the very
important facts

648
00:34:59,340 --> 00:35:02,900
here is, of course, that as you
look around the visual scene,

649
00:35:02,900 --> 00:35:04,820
let's imagine you're
looking at a face.

650
00:35:04,820 --> 00:35:06,392
You're looking at
a face head on.

651
00:35:06,392 --> 00:35:07,975
You're looking at a
face in a profile.

652
00:35:07,975 --> 00:35:09,764
You're looking at a face tilted.

653
00:35:09,764 --> 00:35:11,055
You're looking at a face close.

654
00:35:11,055 --> 00:35:13,200
You're looking at a face far.

655
00:35:13,200 --> 00:35:15,090
And it's still
all the same face.

656
00:35:15,090 --> 00:35:17,920
And yet the
impressions that face

657
00:35:17,920 --> 00:35:23,050
makes on the visual
system, in the retina,

658
00:35:23,050 --> 00:35:27,890
in the geniculate, in the visual
cortex, varies a great deal.

659
00:35:27,890 --> 00:35:31,060
And yet we can come
up with constancy,

660
00:35:31,060 --> 00:35:33,280
which is sort of a
higher level process.

661
00:35:33,280 --> 00:35:36,040
And I'll come to
that in short order.

662
00:35:36,040 --> 00:35:39,180
All right, so now,
therefore, we need

663
00:35:39,180 --> 00:35:44,000
to start to talk about what we
call intermediate level vision.

664
00:35:44,000 --> 00:35:45,470

665
00:35:45,470 --> 00:35:47,350
What is intermediate
level vision?

666
00:35:47,350 --> 00:35:49,760
So far, we've talked
almost exclusively

667
00:35:49,760 --> 00:35:53,470
about basic visual capacities,
color, brightness, pattern,

668
00:35:53,470 --> 00:35:55,720
texture, motion,
depth, the very,

669
00:35:55,720 --> 00:35:57,410
just the very basic types.

670
00:35:57,410 --> 00:36:00,880
But now when we talk about
intermediate visual capacities,

671
00:36:00,880 --> 00:36:02,580
we talk about constancy.

672
00:36:02,580 --> 00:36:04,350
One example of
constancy, of course,

673
00:36:04,350 --> 00:36:06,590
is that the face that
you're looking at,

674
00:36:06,590 --> 00:36:10,010
whether it's profile, or head
on, or near, or far, it's

675
00:36:10,010 --> 00:36:11,140
still the same face.

676
00:36:11,140 --> 00:36:13,810
So we get constancy out of it.

677
00:36:13,810 --> 00:36:16,610
Then there's an
important necessity

678
00:36:16,610 --> 00:36:19,020
to be able to select
various aspects out

679
00:36:19,020 --> 00:36:23,270
of the visual scene, to be
able to select if you recognize

680
00:36:23,270 --> 00:36:28,500
things, to induce
transposition and variance,

681
00:36:28,500 --> 00:36:30,950
and to be able to
make comparisons,

682
00:36:30,950 --> 00:36:34,860
and also lastly, to be able to
say where things are in space.

683
00:36:34,860 --> 00:36:37,190
So those are these
intermediate capacities.

684
00:36:37,190 --> 00:36:40,150
And we'll talk a little
bit about them next.

685
00:36:40,150 --> 00:36:44,650
All right, so here's an
example of constancy.

686
00:36:44,650 --> 00:36:48,715
What we have here
is a bunch of words

687
00:36:48,715 --> 00:36:52,360
for doubt in different
sizes, and orientations,

688
00:36:52,360 --> 00:36:56,020
and some handwritten, and so on.

689
00:36:56,020 --> 00:36:58,260
And yet all of them say
the same thing, doubt.

690
00:36:58,260 --> 00:37:00,210
We extract that, OK.

691
00:37:00,210 --> 00:37:04,720
Now to make this
even more difficult,

692
00:37:04,720 --> 00:37:08,380
I'm going to tell you that
there's one of these words here

693
00:37:08,380 --> 00:37:10,780
which is not doubt
and see who can

694
00:37:10,780 --> 00:37:12,610
find the word that is not doubt.

695
00:37:12,610 --> 00:37:16,380

696
00:37:16,380 --> 00:37:18,840
Everybody find it yet?

697
00:37:18,840 --> 00:37:24,250
OK, right there, not
doubt but doubts.

698
00:37:24,250 --> 00:37:27,570
And so we are able to
extract the difference

699
00:37:27,570 --> 00:37:29,750
in this kind of
visual scene, even

700
00:37:29,750 --> 00:37:31,810
though it is incredibly subtle.

701
00:37:31,810 --> 00:37:35,680
And yet, we are also able
to say that this doubt,

702
00:37:35,680 --> 00:37:38,760
and this doubt, and this
doubt are all the same,

703
00:37:38,760 --> 00:37:42,930
even though they're different
size or different in print.

704
00:37:42,930 --> 00:37:44,890

705
00:37:44,890 --> 00:37:48,220
We can extract from
that the common element.

706
00:37:48,220 --> 00:37:50,160
So this is an
incredible capacity

707
00:37:50,160 --> 00:37:51,440
on part of the visual system.

708
00:37:51,440 --> 00:37:52,500

709
00:37:52,500 --> 00:37:57,570
And of course, the big question
comes up, how is this achieved?

710
00:37:57,570 --> 00:38:00,720
And at this stage, we only
have very preliminary answers

711
00:38:00,720 --> 00:38:01,640
to that.

712
00:38:01,640 --> 00:38:06,630
OK, so now I want to show you a
couple other cute things here.

713
00:38:06,630 --> 00:38:10,035
This is another famous
artist, Hirschfeld,

714
00:38:10,035 --> 00:38:12,220
He is no longer with
us, unfortunately.

715
00:38:12,220 --> 00:38:17,410
He published hundreds
of these little cartoons

716
00:38:17,410 --> 00:38:21,220
in The New Yorker, all right.

717
00:38:21,220 --> 00:38:25,146
Now, you see his named
here, Hirschfeld.

718
00:38:25,146 --> 00:38:26,770
What do you see at
the end of his name?

719
00:38:26,770 --> 00:38:28,620

720
00:38:28,620 --> 00:38:30,410
What's that-- what is this?

721
00:38:30,410 --> 00:38:31,680

722
00:38:31,680 --> 00:38:32,830
What, that?

723
00:38:32,830 --> 00:38:35,085

724
00:38:35,085 --> 00:38:36,510
Huh?

725
00:38:36,510 --> 00:38:37,390
A number three.

726
00:38:37,390 --> 00:38:38,420

727
00:38:38,420 --> 00:38:40,130
Why does it say
Hirschfeld three?

728
00:38:40,130 --> 00:38:42,280
His name isn't Hirschfeld three.

729
00:38:42,280 --> 00:38:44,800
And then if you've
seen several of these,

730
00:38:44,800 --> 00:38:47,170
how many of you have
seen these before?

731
00:38:47,170 --> 00:38:50,530
Oh, you missed out on
something very good.

732
00:38:50,530 --> 00:38:55,290
At least look it up sometimes
again on the internet.

733
00:38:55,290 --> 00:38:57,300
Just type in Hirschfeld.

734
00:38:57,300 --> 00:38:59,650
And some of these will come up.

735
00:38:59,650 --> 00:39:01,060
So he says three.

736
00:39:01,060 --> 00:39:03,300
And some of his
pictures may have one.

737
00:39:03,300 --> 00:39:07,160
Some of his pictures may
have five or six at the end.

738
00:39:07,160 --> 00:39:09,500
And you say, well,
what on earth is that?

739
00:39:09,500 --> 00:39:10,260
Anybody know?

740
00:39:10,260 --> 00:39:11,310

741
00:39:11,310 --> 00:39:11,810
Yes.

742
00:39:11,810 --> 00:39:14,268
AUDIENCE: Doesn't he have other
names hidden in the images?

743
00:39:14,268 --> 00:39:15,610
So maybe that's the [INAUDIBLE].

744
00:39:15,610 --> 00:39:17,234
PROFESSOR: Ah ha,
you're thinking well.

745
00:39:17,234 --> 00:39:18,560
OK, that's getting there.

746
00:39:18,560 --> 00:39:21,355
OK, so he's telling us
something that there

747
00:39:21,355 --> 00:39:23,330
are three of in this picture.

748
00:39:23,330 --> 00:39:24,700
That's what he's telling us.

749
00:39:24,700 --> 00:39:25,530
OK.

750
00:39:25,530 --> 00:39:27,790
So what is the three of
that he has in this picture?

751
00:39:27,790 --> 00:39:29,510

752
00:39:29,510 --> 00:39:31,420
Ah ha, anybody want
to come up with?

753
00:39:31,420 --> 00:39:33,150
It

754
00:39:33,150 --> 00:39:34,650
AUDIENCE: The line orientation?

755
00:39:34,650 --> 00:39:35,300
PROFESSOR: Huh?

756
00:39:35,300 --> 00:39:36,750
AUDIENCE: Like the
orientation of the line?

757
00:39:36,750 --> 00:39:37,708
PROFESSOR: Orientation?

758
00:39:37,708 --> 00:39:38,925

759
00:39:38,925 --> 00:39:39,424
No.

760
00:39:39,424 --> 00:39:41,760

761
00:39:41,760 --> 00:39:42,260
Mhm.

762
00:39:42,260 --> 00:39:44,339

763
00:39:44,339 --> 00:39:45,380
OK, I'll give you a clue.

764
00:39:45,380 --> 00:39:48,796

765
00:39:48,796 --> 00:39:49,920
Hirschfeld had a daughter--

766
00:39:49,920 --> 00:39:51,505

767
00:39:51,505 --> 00:39:52,130
AUDIENCE: Nina.

768
00:39:52,130 --> 00:39:56,990
PROFESSOR: Who he was very
fond-- hey, fantastic, Nina.

769
00:39:56,990 --> 00:39:57,490
OK.

770
00:39:57,490 --> 00:39:58,580

771
00:39:58,580 --> 00:40:00,250
So had a daughter, Nina.

772
00:40:00,250 --> 00:40:08,860
And he decided in his display
to put her name up there.

773
00:40:08,860 --> 00:40:11,700
And now he said three,
that he put her name up

774
00:40:11,700 --> 00:40:13,180
there three times.

775
00:40:13,180 --> 00:40:15,656
Now who can find
all three of them?

776
00:40:15,656 --> 00:40:17,530
AUDIENCE: [INAUDIBLE]
the arm [INAUDIBLE].

777
00:40:17,530 --> 00:40:19,460
PROFESSOR: OK, here's one.

778
00:40:19,460 --> 00:40:21,320
Here is another one.

779
00:40:21,320 --> 00:40:24,330
OK, and here is the
third one, in the arms.

780
00:40:24,330 --> 00:40:26,120

781
00:40:26,120 --> 00:40:29,100
And so if you, that's one of
the fun things you can do.

782
00:40:29,100 --> 00:40:31,480

783
00:40:31,480 --> 00:40:33,530
Back when he was
doing this actively,

784
00:40:33,530 --> 00:40:35,580
every time he had a
cartoon like this,

785
00:40:35,580 --> 00:40:38,775
I would spend some time trying
to see where were the Ninas.

786
00:40:38,775 --> 00:40:40,570
[LAUGHTER]

787
00:40:40,570 --> 00:40:44,350
So that I think is
quite remarkable

788
00:40:44,350 --> 00:40:49,720
and further highlights
the complexities

789
00:40:49,720 --> 00:40:53,830
and the remarkability
of our visual system.

790
00:40:53,830 --> 00:40:56,700
OK, so now, I'm showing
you yet another picture.

791
00:40:56,700 --> 00:40:58,344

792
00:40:58,344 --> 00:40:59,760
In this case again,
you have to be

793
00:40:59,760 --> 00:41:01,270
knowledgeable about something.

794
00:41:01,270 --> 00:41:03,050
Does anybody know who this is?

795
00:41:03,050 --> 00:41:07,090

796
00:41:07,090 --> 00:41:08,162
Oh, all right.

797
00:41:08,162 --> 00:41:09,745
Well, let me see if
I make it smaller.

798
00:41:09,745 --> 00:41:11,170

799
00:41:11,170 --> 00:41:13,010
Can you see that now?

800
00:41:13,010 --> 00:41:14,050
Who is that?

801
00:41:14,050 --> 00:41:16,180
OK, that's Voltaire.

802
00:41:16,180 --> 00:41:17,080
All right.

803
00:41:17,080 --> 00:41:25,090
Now Voltaire, 18th century, is
a guy who had outdone all of us.

804
00:41:25,090 --> 00:41:29,430
And he had written 2,000 books.

805
00:41:29,430 --> 00:41:30,620
Can you believe that?

806
00:41:30,620 --> 00:41:33,280
The guy had written 2,000 books.

807
00:41:33,280 --> 00:41:34,880
But that's not the
prime point here.

808
00:41:34,880 --> 00:41:38,190
Most of us who know a
little bit about history

809
00:41:38,190 --> 00:41:41,560
immediately recognize him
as, oh yeah, that's Voltaire.

810
00:41:41,560 --> 00:41:45,640
But then if you go back here and
you look at this more closely,

811
00:41:45,640 --> 00:41:49,060
and so what I'm going
to do here is just

812
00:41:49,060 --> 00:41:51,170
to tell you what a
remarkable artist.

813
00:41:51,170 --> 00:41:53,350
The guy who did this
is Salvador Dali.

814
00:41:53,350 --> 00:41:56,140
This, by the way,
is his wife, OK.

815
00:41:56,140 --> 00:42:01,090
And so what I'm
going to do here,

816
00:42:01,090 --> 00:42:04,200
I'm going to blow up the
center portion of this figure.

817
00:42:04,200 --> 00:42:08,450

818
00:42:08,450 --> 00:42:09,140
There it is.

819
00:42:09,140 --> 00:42:10,160

820
00:42:10,160 --> 00:42:11,100
Can you see this?

821
00:42:11,100 --> 00:42:13,560
This is two faces, two nuns.

822
00:42:13,560 --> 00:42:15,100
And this is their outfit.

823
00:42:15,100 --> 00:42:16,320

824
00:42:16,320 --> 00:42:20,200
So there's nothing there
that is really Voltaire.

825
00:42:20,200 --> 00:42:21,230

826
00:42:21,230 --> 00:42:23,300
But he's such a
remarkable artist

827
00:42:23,300 --> 00:42:27,900
that he could create Voltaire
by playing around with this, OK.

828
00:42:27,900 --> 00:42:29,990

829
00:42:29,990 --> 00:42:33,035
So let me go back to
the display again.

830
00:42:33,035 --> 00:42:46,900

831
00:42:46,900 --> 00:42:48,920
OK, so this is Salvador Dali.

832
00:42:48,920 --> 00:42:50,200

833
00:42:50,200 --> 00:42:53,720
And certainly, it's an
artist who is remarkable.

834
00:42:53,720 --> 00:42:56,930
He has many paintings
in which this kind

835
00:42:56,930 --> 00:43:01,590
of a double confusing thing,
playing on your ability to see.

836
00:43:01,590 --> 00:43:04,720
So because of that,
for people like me

837
00:43:04,720 --> 00:43:06,890
who's interested in
how we see things,

838
00:43:06,890 --> 00:43:09,830
certainly I'm very
intrigued by his artwork.

839
00:43:09,830 --> 00:43:14,790
And you may enjoy looking
that up again on the internet.

840
00:43:14,790 --> 00:43:21,180
OK, so now I'm going to tell you
yet another interesting aspect

841
00:43:21,180 --> 00:43:23,856
of art involved in this.

842
00:43:23,856 --> 00:43:30,530
And this one has to do with a
book, a very clever book that

843
00:43:30,530 --> 00:43:33,460
was written by David Hockney.

844
00:43:33,460 --> 00:43:35,380
Has anybody ever heard of him?

845
00:43:35,380 --> 00:43:36,720

846
00:43:36,720 --> 00:43:38,810
I'm not surprised since
he's not that well known.

847
00:43:38,810 --> 00:43:39,960

848
00:43:39,960 --> 00:43:43,105
But the book he wrote is
called Secret Knowledge.

849
00:43:43,105 --> 00:43:44,196

850
00:43:44,196 --> 00:43:46,290
Now what was that all about?

851
00:43:46,290 --> 00:43:55,400
Well, he analyzed how artists
created paintings way,

852
00:43:55,400 --> 00:43:59,380
way, way back when and
in subsequent years

853
00:43:59,380 --> 00:44:01,730
up to the present.

854
00:44:01,730 --> 00:44:04,590
And now I'm going to show
you some of his pictures

855
00:44:04,590 --> 00:44:07,040
and give you a sense of
what this is all about.

856
00:44:07,040 --> 00:44:15,490
Here is a picture, OK, by
Masolino da Panicale in 1425.

857
00:44:15,490 --> 00:44:18,530
This was a typical kind
of picture back then.

858
00:44:18,530 --> 00:44:20,780
Artists had a very
poor sense of how

859
00:44:20,780 --> 00:44:24,930
to create depth
perception in a painting.

860
00:44:24,930 --> 00:44:26,440

861
00:44:26,440 --> 00:44:28,840
That was before they came
up with vanishing point.

862
00:44:28,840 --> 00:44:30,210

863
00:44:30,210 --> 00:44:35,430
And so this kind of looks flat
and fairly expressionless.

864
00:44:35,430 --> 00:44:36,690

865
00:44:36,690 --> 00:44:39,220
This was in 1425.

866
00:44:39,220 --> 00:44:44,440
And then Hockney noticed
that just five years later,

867
00:44:44,440 --> 00:44:50,290
another artist came
up with a picture,

868
00:44:50,290 --> 00:44:53,960
just five years later, that
looks almost like a photograph.

869
00:44:53,960 --> 00:44:55,370

870
00:44:55,370 --> 00:44:57,555
And he said, what on
earth has happened?

871
00:44:57,555 --> 00:44:58,940

872
00:44:58,940 --> 00:45:01,610
And that's why the book is
called the Secret Knowledge.

873
00:45:01,610 --> 00:45:02,560
So what happened?

874
00:45:02,560 --> 00:45:03,200
Anybody know?

875
00:45:03,200 --> 00:45:05,657

876
00:45:05,657 --> 00:45:06,240
Ah, all right.

877
00:45:06,240 --> 00:45:07,739
Well, let me tell
you what happened.

878
00:45:07,739 --> 00:45:16,530
What happened is that at
that time, in the 1400s,

879
00:45:16,530 --> 00:45:19,800
they came up with the lens, OK.

880
00:45:19,800 --> 00:45:21,350

881
00:45:21,350 --> 00:45:27,385
And so they created a device
called the camera obscura.

882
00:45:27,385 --> 00:45:28,710

883
00:45:28,710 --> 00:45:30,670
Does anybody know what
the camera obscura is?

884
00:45:30,670 --> 00:45:32,140

885
00:45:32,140 --> 00:45:39,520
It's essentially very
similar to a camera, OK.

886
00:45:39,520 --> 00:45:40,550

887
00:45:40,550 --> 00:45:41,940
So here we have it.

888
00:45:41,940 --> 00:45:43,120
Here's a camera obscura.

889
00:45:43,120 --> 00:45:44,380

890
00:45:44,380 --> 00:45:47,240
And what they did,
these artists,

891
00:45:47,240 --> 00:45:49,580
they would create a building.

892
00:45:49,580 --> 00:45:51,400
So it will be dark inside.

893
00:45:51,400 --> 00:45:53,280
And then they would
put the lens here.

894
00:45:53,280 --> 00:45:55,110
And they would put
some object out here

895
00:45:55,110 --> 00:45:59,390
that they wanted to
make a painting of.

896
00:45:59,390 --> 00:46:04,670
And that would be reflected
onto a piece of canvas here.

897
00:46:04,670 --> 00:46:06,890
Of course, it would
be upside down, right?

898
00:46:06,890 --> 00:46:09,050
And then they would paint it.

899
00:46:09,050 --> 00:46:12,980
And so once it was painted,
they could turn it around

900
00:46:12,980 --> 00:46:16,910
and finish it and sell it.

901
00:46:16,910 --> 00:46:20,040
Now, the reason they did this,
the prime reason they did this,

902
00:46:20,040 --> 00:46:23,470
is because it was much
much, much quicker

903
00:46:23,470 --> 00:46:25,700
to create a portrait,
for example,

904
00:46:25,700 --> 00:46:29,880
by using this procedure than
to actually look at a person

905
00:46:29,880 --> 00:46:35,620
and paint them on the
canvas while looking at it.

906
00:46:35,620 --> 00:46:36,120
OK.

907
00:46:36,120 --> 00:46:37,230

908
00:46:37,230 --> 00:46:38,720
So that was done.

909
00:46:38,720 --> 00:46:41,270
And of course, this was
sort of a no, no thing.

910
00:46:41,270 --> 00:46:44,850
And because of that,
it was kept secret.

911
00:46:44,850 --> 00:46:47,780
All these artists who did this,
and some very famous artists

912
00:46:47,780 --> 00:46:57,540
did so, were very careful
never to disclose to the public

913
00:46:57,540 --> 00:46:58,950
that they did this
kind of thing.

914
00:46:58,950 --> 00:47:02,780
Because it was conceived to be
kind of a cheating thing, OK.

915
00:47:02,780 --> 00:47:08,390
So what happened then was
that all these paintings were

916
00:47:08,390 --> 00:47:09,170
created.

917
00:47:09,170 --> 00:47:13,100
And here's an example
by van Eyck in 1436.

918
00:47:13,100 --> 00:47:15,430
Again, this looks much
like a photograph.

919
00:47:15,430 --> 00:47:18,210
But some other thing
is a bit distorted.

920
00:47:18,210 --> 00:47:24,050
So Hockney undertook a
careful, detailed analysis

921
00:47:24,050 --> 00:47:26,810
of how we could tell
whether a painting was

922
00:47:26,810 --> 00:47:31,380
a real painting or one that
used the camera obscura method.

923
00:47:31,380 --> 00:47:33,220
And that's also a real
painting in a sense,

924
00:47:33,220 --> 00:47:37,270
but a painting using the
camera obscura method.

925
00:47:37,270 --> 00:47:39,580
And he came up with
a series of criteria

926
00:47:39,580 --> 00:47:40,850
which are listed in the book.

927
00:47:40,850 --> 00:47:43,140
But I'm only going to
deal with one of them.

928
00:47:43,140 --> 00:47:44,535
OK, so here is an example.

929
00:47:44,535 --> 00:47:47,050

930
00:47:47,050 --> 00:47:49,510
This is in 1597.

931
00:47:49,510 --> 00:47:51,473
This is by Caravaggio, OK.

932
00:47:51,473 --> 00:47:52,830

933
00:47:52,830 --> 00:47:54,540
And what is notable
about this person?

934
00:47:54,540 --> 00:47:59,940

935
00:47:59,940 --> 00:48:03,580
Well, to Hockney,
what was notable

936
00:48:03,580 --> 00:48:05,820
is that this person
is holding the wine

937
00:48:05,820 --> 00:48:07,684
glass in his left hand.

938
00:48:07,684 --> 00:48:08,184
Yeah?

939
00:48:08,184 --> 00:48:09,440

940
00:48:09,440 --> 00:48:11,000
And he said, huh.

941
00:48:11,000 --> 00:48:12,069
That's curious.

942
00:48:12,069 --> 00:48:14,360
And then he looked at a whole
bunch of other paintings.

943
00:48:14,360 --> 00:48:15,810
And he had one in
which there were

944
00:48:15,810 --> 00:48:19,090
three people on the painting.

945
00:48:19,090 --> 00:48:21,170
And all three of them
were left handed.

946
00:48:21,170 --> 00:48:22,350
Yeah.

947
00:48:22,350 --> 00:48:24,760
And he said, my god,
that is really curious.

948
00:48:24,760 --> 00:48:26,200
And he said, well,
let me analyze

949
00:48:26,200 --> 00:48:30,480
what happens when you
do this kind of stuff

950
00:48:30,480 --> 00:48:33,000
with the camera obscura method.

951
00:48:33,000 --> 00:48:34,780
So here's the example of this.

952
00:48:34,780 --> 00:48:38,350
This is the original
image, he claims,

953
00:48:38,350 --> 00:48:41,630
meaning the person is right
handed, is not left handed.

954
00:48:41,630 --> 00:48:45,470
Then you put this
person through the lens

955
00:48:45,470 --> 00:48:49,940
and put him up here
upside down like that, OK.

956
00:48:49,940 --> 00:48:53,520
He's painted on the
canvas like that.

957
00:48:53,520 --> 00:48:58,910
And then what you do is you
rotate this 180 degrees.

958
00:48:58,910 --> 00:49:00,930
And when you do
that, lo and behold,

959
00:49:00,930 --> 00:49:02,660
the person becomes left hand.

960
00:49:02,660 --> 00:49:05,110
And to make this clear,
I added the F here.

961
00:49:05,110 --> 00:49:07,360
So this is a normal
F. This is what

962
00:49:07,360 --> 00:49:12,130
is projected onto the canvas.

963
00:49:12,130 --> 00:49:14,370
And this is when you
rotate it 180 degrees.

964
00:49:14,370 --> 00:49:15,540

965
00:49:15,540 --> 00:49:19,370
So you reverse the left
according to Hockney,

966
00:49:19,370 --> 00:49:22,130
reverse the image.

967
00:49:22,130 --> 00:49:29,300
And so it was a dead giveaway
that most of the people who

968
00:49:29,300 --> 00:49:33,360
appeared left handed in his
paintings, in these paintings,

969
00:49:33,360 --> 00:49:34,910
I should say, not
just Caravaggio,

970
00:49:34,910 --> 00:49:39,910
but several other people,
used the camera obscura method

971
00:49:39,910 --> 00:49:42,510
for creating the painting.

972
00:49:42,510 --> 00:49:46,680
All right, so that
is the process.

973
00:49:46,680 --> 00:49:50,250
And then I want to show you
one more example of this.

974
00:49:50,250 --> 00:49:53,820
OK, this is a very famous
painting also, all right.

975
00:49:53,820 --> 00:49:55,320

976
00:49:55,320 --> 00:50:00,576
This is the so-called marriage
of Giovanni Arnolfini.

977
00:50:00,576 --> 00:50:03,070

978
00:50:03,070 --> 00:50:08,370
And this by van Eyck, again
1493, way, way, way, way, way,

979
00:50:08,370 --> 00:50:09,280
way back then.

980
00:50:09,280 --> 00:50:11,656
Now this is a famous painting.

981
00:50:11,656 --> 00:50:13,030
Now there's one
interesting thing

982
00:50:13,030 --> 00:50:14,080
first of all I
want to point out.

983
00:50:14,080 --> 00:50:15,100
You see this here?

984
00:50:15,100 --> 00:50:16,670

985
00:50:16,670 --> 00:50:19,080
That's proof that
in those days they

986
00:50:19,080 --> 00:50:22,150
had come up with the lens, yeah.

987
00:50:22,150 --> 00:50:24,700

988
00:50:24,700 --> 00:50:28,100
Now what is wrong
with this picture?

989
00:50:28,100 --> 00:50:32,100

990
00:50:32,100 --> 00:50:35,700
This guy who is about
to marry this woman

991
00:50:35,700 --> 00:50:38,380
is holding her
with his left hand.

992
00:50:38,380 --> 00:50:39,114

993
00:50:39,114 --> 00:50:40,280
I mean, that's unacceptable.

994
00:50:40,280 --> 00:50:41,840

995
00:50:41,840 --> 00:50:45,010
You're supposed to hold it
with the right hand, yeah.

996
00:50:45,010 --> 00:50:47,400
Now the reason he's holding
it with the left hand

997
00:50:47,400 --> 00:50:54,200
is because the artist, van
Eyck, used the camera, according

998
00:50:54,200 --> 00:51:01,460
to Hockney, used the camera
obscura method to take this,

999
00:51:01,460 --> 00:51:03,710
to paint this picture, OK.

1000
00:51:03,710 --> 00:51:07,420
And then he rotated
180 degrees and became,

1001
00:51:07,420 --> 00:51:08,965
he became left
handed as a result.

1002
00:51:08,965 --> 00:51:11,690

1003
00:51:11,690 --> 00:51:16,040
Now if instead of having
done that, you would go back,

1004
00:51:16,040 --> 00:51:17,530
you don't have to go back.

1005
00:51:17,530 --> 00:51:18,865

1006
00:51:18,865 --> 00:51:21,300
Go back to here, and
let's go back to this.

1007
00:51:21,300 --> 00:51:28,750
If you take this guy here
and you create the picture.

1008
00:51:28,750 --> 00:51:32,190
But then instead of rotating
it, if you could flip it,

1009
00:51:32,190 --> 00:51:35,860
then you would, he will
remain right handed.

1010
00:51:35,860 --> 00:51:39,380
But of course, we can't do
that because it's on a canvas.

1011
00:51:39,380 --> 00:51:40,505
It's not on a transparency.

1012
00:51:40,505 --> 00:51:42,240

1013
00:51:42,240 --> 00:51:48,780
So that then is the
interesting story of artwork

1014
00:51:48,780 --> 00:51:51,820
that was created using
the camera obscura

1015
00:51:51,820 --> 00:51:55,790
system that further
the highlights

1016
00:51:55,790 --> 00:52:03,310
the amazing interestingly
complicated manner

1017
00:52:03,310 --> 00:52:06,620
in which we can analyze the
visual scene for shapes.

1018
00:52:06,620 --> 00:52:07,690

1019
00:52:07,690 --> 00:52:14,100
All right, now another factor
that is in a similar vein

1020
00:52:14,100 --> 00:52:17,250
has to do with the
recognition of faces.

1021
00:52:17,250 --> 00:52:20,360
Lots of experiments have
been done, including several

1022
00:52:20,360 --> 00:52:24,090
in our department here,
that has recognized

1023
00:52:24,090 --> 00:52:32,720
that, unfortunate use of
terms, recognition, recognize,

1024
00:52:32,720 --> 00:52:36,940
became aware of the fact that
facial recognition depends

1025
00:52:36,940 --> 00:52:40,720
very heavily on seeing
faces right side up.

1026
00:52:40,720 --> 00:52:46,350
When faces are upside down,
you have great difficulty

1027
00:52:46,350 --> 00:52:48,720
telling who is who, OK.

1028
00:52:48,720 --> 00:52:50,080
So let's do that.

1029
00:52:50,080 --> 00:52:51,760
Here are a bunch of faces.

1030
00:52:51,760 --> 00:52:55,000
And I bet you can you
can tell this one, right?

1031
00:52:55,000 --> 00:52:55,540
Who is it?

1032
00:52:55,540 --> 00:52:57,250

1033
00:52:57,250 --> 00:52:58,150
And who is this?

1034
00:52:58,150 --> 00:52:59,440

1035
00:52:59,440 --> 00:53:00,000
Very good.

1036
00:53:00,000 --> 00:53:01,840

1037
00:53:01,840 --> 00:53:03,920
And those two you don't know.

1038
00:53:03,920 --> 00:53:06,770
The problem is that I am
going to flip this over now.

1039
00:53:06,770 --> 00:53:08,780
And you still don't
know who those two are.

1040
00:53:08,780 --> 00:53:13,110

1041
00:53:13,110 --> 00:53:16,290
OK, so this one
is Norbert Wiener.

1042
00:53:16,290 --> 00:53:18,185
Now all of you know
about Norbert Wiener.

1043
00:53:18,185 --> 00:53:20,590
He is one of the great
geniuses of our time.

1044
00:53:20,590 --> 00:53:22,415
He came up with
the digital code.

1045
00:53:22,415 --> 00:53:23,600

1046
00:53:23,600 --> 00:53:27,030
He used to be a
professor at MIT.

1047
00:53:27,030 --> 00:53:29,340
And this here is Chuck Vest.

1048
00:53:29,340 --> 00:53:31,240
Anybody recognize him?

1049
00:53:31,240 --> 00:53:32,490

1050
00:53:32,490 --> 00:53:36,650
Chuck Vest was a
president of MIT for what?

1051
00:53:36,650 --> 00:53:39,620
For eight years, I believe,
maybe more, 12 years?

1052
00:53:39,620 --> 00:53:40,120
I forget.

1053
00:53:40,120 --> 00:53:42,200

1054
00:53:42,200 --> 00:53:47,740
And he has sunk into
obscurity, even though he

1055
00:53:47,740 --> 00:53:50,000
was incredibly visible
for many, many years.

1056
00:53:50,000 --> 00:53:53,970
If I had shown this to you
say eight or 10 years ago,

1057
00:53:53,970 --> 00:53:55,980
you would immediately
recognize him.

1058
00:53:55,980 --> 00:53:59,470
Because he, at that time,
was a president at MIT.

1059
00:53:59,470 --> 00:54:00,870

1060
00:54:00,870 --> 00:54:03,890
So that then is an
interesting fact

1061
00:54:03,890 --> 00:54:08,660
that even though we are capable
of using intermediate level

1062
00:54:08,660 --> 00:54:12,250
vision in a very
sophisticated way,

1063
00:54:12,250 --> 00:54:16,430
it does not seem to work that
well for upside down faces.

1064
00:54:16,430 --> 00:54:18,210

1065
00:54:18,210 --> 00:54:20,620
You guys did pretty
good with those.

1066
00:54:20,620 --> 00:54:22,290
But it takes a while.

1067
00:54:22,290 --> 00:54:27,290
If I had flashed that
on in a tachistoscope,

1068
00:54:27,290 --> 00:54:31,260
you wouldn't have had a vague
idea of who those people were.

1069
00:54:31,260 --> 00:54:33,650
But once it's on for a while
and you can analyze it,

1070
00:54:33,650 --> 00:54:38,780
you can eventually tell
even an upside down face.

1071
00:54:38,780 --> 00:54:47,700
OK, now in the same vein
as we are talking about,

1072
00:54:47,700 --> 00:54:57,680
our ability to process shapes
based on contours and whatnot,

1073
00:54:57,680 --> 00:55:01,330
one of the interesting set of
experiments people had done

1074
00:55:01,330 --> 00:55:05,070
is to look at what is
called subjective contours.

1075
00:55:05,070 --> 00:55:07,690
And so let me give you a
couple of examples of that.

1076
00:55:07,690 --> 00:55:09,160
Here's an example.

1077
00:55:09,160 --> 00:55:11,520
Almost instantly, here
you can see a disk.

1078
00:55:11,520 --> 00:55:15,430
And here you can
see a square rotated

1079
00:55:15,430 --> 00:55:18,980
45 degrees, a
diamond if you will.

1080
00:55:18,980 --> 00:55:20,140

1081
00:55:20,140 --> 00:55:22,290
But if you analyze
this carefully,

1082
00:55:22,290 --> 00:55:27,150
you can see that about
80% of this border

1083
00:55:27,150 --> 00:55:28,940
here is not a border.

1084
00:55:28,940 --> 00:55:30,270
There's no border here.

1085
00:55:30,270 --> 00:55:32,830
There's no border here,
here, here, here, here.

1086
00:55:32,830 --> 00:55:34,770
And yet we can see a square.

1087
00:55:34,770 --> 00:55:39,350
So there's some strange ability
in part of the visual system

1088
00:55:39,350 --> 00:55:43,510
to complete inferred borders.

1089
00:55:43,510 --> 00:55:45,950
All right, I'll come back
to that in just a minute.

1090
00:55:45,950 --> 00:55:49,240
Now another example
of this is shown here.

1091
00:55:49,240 --> 00:55:50,740

1092
00:55:50,740 --> 00:55:53,040
Can you make out
what it says here?

1093
00:55:53,040 --> 00:55:54,024

1094
00:55:54,024 --> 00:55:56,190
If you look a little bit,
you should be able to see.

1095
00:55:56,190 --> 00:55:57,760
What is it?

1096
00:55:57,760 --> 00:55:59,000
Visual system, very good.

1097
00:55:59,000 --> 00:56:00,730
It was difficult to see.

1098
00:56:00,730 --> 00:56:04,020
But as soon as I
turn this into color,

1099
00:56:04,020 --> 00:56:05,270
you have no trouble at all.

1100
00:56:05,270 --> 00:56:06,490

1101
00:56:06,490 --> 00:56:09,290
And that highlights
another important reason

1102
00:56:09,290 --> 00:56:13,610
why color vision is so
useful and has evolved.

1103
00:56:13,610 --> 00:56:16,940
Because it enables
us to see borders,

1104
00:56:16,940 --> 00:56:19,980
where under certain conditions
and lighting conditions,

1105
00:56:19,980 --> 00:56:24,380
borders would not be visible
on the black and white.

1106
00:56:24,380 --> 00:56:29,020
OK, so now that's another
interesting example,

1107
00:56:29,020 --> 00:56:33,670
has to do with further
subjective contours.

1108
00:56:33,670 --> 00:56:37,040
And here's an example of
one with a high contrast,

1109
00:56:37,040 --> 00:56:39,600
where you can
readily see a cube.

1110
00:56:39,600 --> 00:56:41,473
Does everybody see a cube there?

1111
00:56:41,473 --> 00:56:45,130
All right, but if you look
here, it's next to impossible

1112
00:56:45,130 --> 00:56:46,120
to see that.

1113
00:56:46,120 --> 00:56:48,730
Because here the
stimuli are isoluminant.

1114
00:56:48,730 --> 00:56:50,820

1115
00:56:50,820 --> 00:56:55,290
So here you eliminated
contrast, which

1116
00:56:55,290 --> 00:56:57,290
is a very important
aspect of being

1117
00:56:57,290 --> 00:57:05,380
able to analyze the visual
scene in various ways,

1118
00:57:05,380 --> 00:57:06,680
including three dimensions.

1119
00:57:06,680 --> 00:57:11,020
All right, so now here's a
very interesting discovery that

1120
00:57:11,020 --> 00:57:14,200
was made by recording
an area, V2.

1121
00:57:14,200 --> 00:57:17,920

1122
00:57:17,920 --> 00:57:20,730
The recordings were
then made to see

1123
00:57:20,730 --> 00:57:25,850
whether those neurons
perceive subjective contours.

1124
00:57:25,850 --> 00:57:27,880
So here's a receptive field.

1125
00:57:27,880 --> 00:57:31,220
And you take this bar and move
it back and forth across it.

1126
00:57:31,220 --> 00:57:33,670
And you can see it gives
a vigorous response.

1127
00:57:33,670 --> 00:57:35,710
Now you do the same thing.

1128
00:57:35,710 --> 00:57:39,760
But you don't, here you have
only a subjective contour.

1129
00:57:39,760 --> 00:57:42,500
And then you move this
back and forth across.

1130
00:57:42,500 --> 00:57:49,340
Somehow the information is
added up from other areas,

1131
00:57:49,340 --> 00:57:53,940
so that this cell
responds, not that well,

1132
00:57:53,940 --> 00:57:58,130
but responds reasonably well
to the subjective contour.

1133
00:57:58,130 --> 00:58:02,440
So it's said that in
V2, you can carry out

1134
00:58:02,440 --> 00:58:05,440
some of these higher
level processes that

1135
00:58:05,440 --> 00:58:09,200
enables you to
complete figures even

1136
00:58:09,200 --> 00:58:11,890
when they are
incomplete like that.

1137
00:58:11,890 --> 00:58:14,290
OK, here's another
example of that.

1138
00:58:14,290 --> 00:58:15,970
This is even more dramatic.

1139
00:58:15,970 --> 00:58:18,600
In this case, we take this
bar back and forth across.

1140
00:58:18,600 --> 00:58:20,959
That's a vigorous response.

1141
00:58:20,959 --> 00:58:22,250
And then you do the same thing.

1142
00:58:22,250 --> 00:58:27,370
You create a bar here simply
by these continuous horizontal

1143
00:58:27,370 --> 00:58:28,420
bars.

1144
00:58:28,420 --> 00:58:30,980
And still, the cell
responds quite well,

1145
00:58:30,980 --> 00:58:34,090
as if it were
seeing an edge here.

1146
00:58:34,090 --> 00:58:38,140
So we have this kind of
completion in area V2.

1147
00:58:38,140 --> 00:58:40,290
So this is sort
of an initial hint

1148
00:58:40,290 --> 00:58:44,650
then that already
in area V2, you

1149
00:58:44,650 --> 00:58:50,210
begin to process
higher level events,

1150
00:58:50,210 --> 00:58:52,830
among which is the
fact that you can

1151
00:58:52,830 --> 00:58:54,690
complete incomplete contours.

1152
00:58:54,690 --> 00:58:55,780

1153
00:58:55,780 --> 00:58:59,440
All right, so now, the next
thing we're going to turn to,

1154
00:58:59,440 --> 00:59:01,970
we're going to ask
the question, when

1155
00:59:01,970 --> 00:59:05,740
we deal with these
so-called intermediate level

1156
00:59:05,740 --> 00:59:11,900
visual capacities, what
happens when you take out

1157
00:59:11,900 --> 00:59:16,850
such areas as V4 and
MT to your ability

1158
00:59:16,850 --> 00:59:19,710
to see these intermediate
visual capacities?

1159
00:59:19,710 --> 00:59:21,990
So let me describe
to you some of these,

1160
00:59:21,990 --> 00:59:25,230
how one would do
experiments like this.

1161
00:59:25,230 --> 00:59:27,680
First of all,
monkeys are trained.

1162
00:59:27,680 --> 00:59:30,320
This is done with monkeys, of
course, because we can't just

1163
00:59:30,320 --> 00:59:32,900
take a human and remove V4 MT.

1164
00:59:32,900 --> 00:59:38,820
And so what you do here is you
first present a fixation spot.

1165
00:59:38,820 --> 00:59:43,390
Once the monkey fixates,
you present a shape,

1166
00:59:43,390 --> 00:59:45,060
in this case a square.

1167
00:59:45,060 --> 00:59:49,300
And then you present a whole
bunch of other ones, only one

1168
00:59:49,300 --> 00:59:51,350
of which is the same
as the original.

1169
00:59:51,350 --> 00:59:53,450
And the monkey has to
make a [INAUDIBLE] there

1170
00:59:53,450 --> 00:59:54,710
to be rewarded.

1171
00:59:54,710 --> 00:59:59,090
So he has to be able to
detect identity, which

1172
00:59:59,090 --> 01:00:01,850
is an intermediate
visual capacity.

1173
01:00:01,850 --> 01:00:03,850
So that's what this is like.

1174
01:00:03,850 --> 01:00:08,180
And so then what
you can do here is

1175
01:00:08,180 --> 01:00:10,930
to vary the amount
of information

1176
01:00:10,930 --> 01:00:15,740
you can provide, again by
reducing the amount of contour

1177
01:00:15,740 --> 01:00:17,650
information that
you can provide.

1178
01:00:17,650 --> 01:00:19,120
You can do this in several ways.

1179
01:00:19,120 --> 01:00:21,320

1180
01:00:21,320 --> 01:00:25,020
First of all, before we do that,
I'll come to that in a second,

1181
01:00:25,020 --> 01:00:26,770
let's just see, how
does the monkey do

1182
01:00:26,770 --> 01:00:30,040
when you do those very similar
shapes I just have shown you

1183
01:00:30,040 --> 01:00:32,870
after you take out area V4?

1184
01:00:32,870 --> 01:00:36,370
What you find here is that the
monkey initially, after you

1185
01:00:36,370 --> 01:00:42,030
take out the area V4, can't
even do the regular task, which

1186
01:00:42,030 --> 01:00:43,530
is this one, right?

1187
01:00:43,530 --> 01:00:46,140
So this is identical to
this one in this case.

1188
01:00:46,140 --> 01:00:49,540
And when you do that, he does
very poorly to begin with

1189
01:00:49,540 --> 01:00:51,710
and then gradually,
over many, many days,

1190
01:00:51,710 --> 01:00:53,540
improves a great deal.

1191
01:00:53,540 --> 01:00:56,917
Then if you put a new figure
in, then it takes a while for it

1192
01:00:56,917 --> 01:00:59,070
to learn that, all right.

1193
01:00:59,070 --> 01:01:02,520
Now then, what you
can do, now I come

1194
01:01:02,520 --> 01:01:06,550
to this business of
having these same figures.

1195
01:01:06,550 --> 01:01:09,720
But you can vary
them now, so that you

1196
01:01:09,720 --> 01:01:13,450
have to do a transposition to
do an intermediate visual task.

1197
01:01:13,450 --> 01:01:17,760
In this case, what you do
is that you vary the size.

1198
01:01:17,760 --> 01:01:20,200
In this case, it's
identical just like it was.

1199
01:01:20,200 --> 01:01:23,394
In this case, you see that
this is smaller than this.

1200
01:01:23,394 --> 01:01:25,310
But you say, oh, I have
to go find the circle.

1201
01:01:25,310 --> 01:01:26,530

1202
01:01:26,530 --> 01:01:27,960
I'm not looking for identity.

1203
01:01:27,960 --> 01:01:31,050
I'm looking for something
that's the same looking.

1204
01:01:31,050 --> 01:01:32,680
And here we have
even bigger case.

1205
01:01:32,680 --> 01:01:33,749
This is a triangle.

1206
01:01:33,749 --> 01:01:35,290
And so the monkey
makes a [INAUDIBLE]

1207
01:01:35,290 --> 01:01:36,940
to that one here to this one.

1208
01:01:36,940 --> 01:01:40,870
So that has to do a
test position in size.

1209
01:01:40,870 --> 01:01:43,350
Another thing you can
do that we talked about,

1210
01:01:43,350 --> 01:01:46,060
you can vary the amount
of contour information

1211
01:01:46,060 --> 01:01:48,480
by doing this kind of occlusion.

1212
01:01:48,480 --> 01:01:50,130
And the degree of
occlusion you can

1213
01:01:50,130 --> 01:01:53,530
vary by varying the spatial
frequency of the display.

1214
01:01:53,530 --> 01:01:57,500
And lastly, you can also
decrease or increase the amount

1215
01:01:57,500 --> 01:02:00,170
of contour information
that you provide.

1216
01:02:00,170 --> 01:02:05,760
Now if you do with this, you get
a huge effect after V4 lesion.

1217
01:02:05,760 --> 01:02:07,210

1218
01:02:07,210 --> 01:02:11,390
This was the normal condition
with the varied object size.

1219
01:02:11,390 --> 01:02:13,040
This is the occlusion.

1220
01:02:13,040 --> 01:02:16,530
And here is the varied
contour information.

1221
01:02:16,530 --> 01:02:19,020
You can see that there is
quite a dramatic loss, not

1222
01:02:19,020 --> 01:02:23,600
a total, but quite a notable
loss, in the monkey's ability

1223
01:02:23,600 --> 01:02:25,130
to perform this task.

1224
01:02:25,130 --> 01:02:28,010
And this is also reflected
in the huge increase

1225
01:02:28,010 --> 01:02:32,230
in the latencies with which the
monkey can perform the task.

1226
01:02:32,230 --> 01:02:35,720
So area V4 seems to
play an important role

1227
01:02:35,720 --> 01:02:38,860
in these intermediate
visual capacities.

1228
01:02:38,860 --> 01:02:42,240
And I come to some more examples
of that in just a minute.

1229
01:02:42,240 --> 01:02:47,150
Now, yet another
important factor

1230
01:02:47,150 --> 01:02:59,400
in analyzing the
visual scene occurs

1231
01:02:59,400 --> 01:03:02,530
when it is our task
to find something

1232
01:03:02,530 --> 01:03:05,540
out there that is
less noticeable.

1233
01:03:05,540 --> 01:03:06,780

1234
01:03:06,780 --> 01:03:13,830
Remember, we talked a little
bit about using camouflage,

1235
01:03:13,830 --> 01:03:18,105
in which case, you have to find
something lesser to survive.

1236
01:03:18,105 --> 01:03:19,300

1237
01:03:19,300 --> 01:03:22,520
Now in this case,
what we can do here,

1238
01:03:22,520 --> 01:03:24,740
do a similar
experiment in a monkey.

1239
01:03:24,740 --> 01:03:26,130

1240
01:03:26,130 --> 01:03:28,820
On the left side,
the target, the one

1241
01:03:28,820 --> 01:03:32,330
the monkey's supposed to select,
has a much higher contrast

1242
01:03:32,330 --> 01:03:34,660
than the distractors.

1243
01:03:34,660 --> 01:03:36,610
And you can vary the
degree of difference,

1244
01:03:36,610 --> 01:03:39,760
but always the target is
brighter than the others,

1245
01:03:39,760 --> 01:03:41,240
so that it stands out.

1246
01:03:41,240 --> 01:03:44,740
But then, you must
be equally able,

1247
01:03:44,740 --> 01:03:46,280
whether you're an
animal or a human,

1248
01:03:46,280 --> 01:03:47,990
to be able to pull
out something that's

1249
01:03:47,990 --> 01:03:50,615
lesser in the visuals field.

1250
01:03:50,615 --> 01:03:51,720

1251
01:03:51,720 --> 01:03:55,120
And you have to
be able to do this

1252
01:03:55,120 --> 01:04:02,555
if you are going to survive
in your environment.

1253
01:04:02,555 --> 01:04:06,120
Now here then, the task is to
go to this lesser stimulus,

1254
01:04:06,120 --> 01:04:07,780
because it's the odd stimulus.

1255
01:04:07,780 --> 01:04:10,730
So what you're extracting
is the odd stimulus.

1256
01:04:10,730 --> 01:04:14,270
Here it's easy to extract, and
here it's difficult to extract.

1257
01:04:14,270 --> 01:04:21,830
And so now the question is what
happens in the visual cortex?

1258
01:04:21,830 --> 01:04:25,230
What area plays a role in this?

1259
01:04:25,230 --> 01:04:28,450
And so it was
discovered that area V4

1260
01:04:28,450 --> 01:04:29,690
is very important for this.

1261
01:04:29,690 --> 01:04:31,600
And let me explain
this to you then.

1262
01:04:31,600 --> 01:04:36,480
You do an experiment in
which you remove area V4

1263
01:04:36,480 --> 01:04:37,520
and see what happens.

1264
01:04:37,520 --> 01:04:39,740

1265
01:04:39,740 --> 01:04:42,900
And here what we have is
when the star gets brighter,

1266
01:04:42,900 --> 01:04:45,940
you vary the
luminance difference.

1267
01:04:45,940 --> 01:04:48,900
And you can see there's a mild
deficit with the V4 lesion,

1268
01:04:48,900 --> 01:04:50,980
highly significant
but still fairly mild.

1269
01:04:50,980 --> 01:04:53,760
On the other hand, when
you make it dimmer,

1270
01:04:53,760 --> 01:04:57,800
the monkey is practically
staying at the probability.

1271
01:04:57,800 --> 01:04:59,050

1272
01:04:59,050 --> 01:05:01,190
He cannot do the task at all.

1273
01:05:01,190 --> 01:05:04,360
So somehow, V4 plays
a very important role

1274
01:05:04,360 --> 01:05:09,200
in being able to ferret
out some subtle things

1275
01:05:09,200 --> 01:05:11,050
in the environment,
lesser things.

1276
01:05:11,050 --> 01:05:12,410

1277
01:05:12,410 --> 01:05:15,860
All right, so this
then brings us

1278
01:05:15,860 --> 01:05:17,970
to yet another way
of analyzing this.

1279
01:05:17,970 --> 01:05:21,370
We can, in this case,
go to the larger target,

1280
01:05:21,370 --> 01:05:23,250
in this case to
the smaller target.

1281
01:05:23,250 --> 01:05:25,660
And you can recognize yourself
that this is certainly

1282
01:05:25,660 --> 01:05:27,190
easier than that.

1283
01:05:27,190 --> 01:05:31,820
But a normal monkey can do
both of these quite well,

1284
01:05:31,820 --> 01:05:32,750
shown here.

1285
01:05:32,750 --> 01:05:34,520
This is the normal
monkey's performance,

1286
01:05:34,520 --> 01:05:37,300
the target larger, here
the target smaller.

1287
01:05:37,300 --> 01:05:38,840
He does extremely well.

1288
01:05:38,840 --> 01:05:41,880
But with a V4 lesion when
the target is smaller,

1289
01:05:41,880 --> 01:05:50,150
it's totally devastated, so
as if area V4 were involved

1290
01:05:50,150 --> 01:05:57,360
in the analysis of subtle
things, things which are lesser

1291
01:05:57,360 --> 01:05:59,750
then, rather than
being reflex like

1292
01:05:59,750 --> 01:06:04,640
and going to the brightest,
biggest thing in the world.

1293
01:06:04,640 --> 01:06:12,270
OK, so that then indeed
highlights the fact

1294
01:06:12,270 --> 01:06:13,970
that we have some
of these areas,

1295
01:06:13,970 --> 01:06:17,850
including V4 and of
course, therefore MT,

1296
01:06:17,850 --> 01:06:25,030
that involved in these much more
subtle types of visual analyses

1297
01:06:25,030 --> 01:06:26,920
that we need to perform.

1298
01:06:26,920 --> 01:06:34,350
Now capitalizing on these
kinds of subtle things

1299
01:06:34,350 --> 01:06:37,670
that we are capable of doing,
artists, in addition to

1300
01:06:37,670 --> 01:06:41,160
the ones that I've
shown you before, also

1301
01:06:41,160 --> 01:06:48,210
created all kinds of percepts, I
should say paintings, sketches,

1302
01:06:48,210 --> 01:06:54,730
precepts, that cause
confusion by playing around

1303
01:06:54,730 --> 01:06:55,950
with these factors.

1304
01:06:55,950 --> 01:07:01,345
One of these here is a very well
known audience called Escher.

1305
01:07:01,345 --> 01:07:03,350

1306
01:07:03,350 --> 01:07:07,940
This was so near the
end of the 19th century.

1307
01:07:07,940 --> 01:07:11,155
And here what we
have is this is sort

1308
01:07:11,155 --> 01:07:13,740
of a figure-ground confusion.

1309
01:07:13,740 --> 01:07:17,400
Here what we have is a bunch
of birds that fly to the right,

1310
01:07:17,400 --> 01:07:20,060
and also a bunch of birds
that fly to the left.

1311
01:07:20,060 --> 01:07:21,820
And so it's confusing.

1312
01:07:21,820 --> 01:07:22,760
It's alternating.

1313
01:07:22,760 --> 01:07:24,420
You don't know
which one is which.

1314
01:07:24,420 --> 01:07:27,960
And it has to do with a
very, very clever creation

1315
01:07:27,960 --> 01:07:31,370
of figure-ground confusions.

1316
01:07:31,370 --> 01:07:32,700

1317
01:07:32,700 --> 01:07:36,220
And actually, next time
we talk about illusions,

1318
01:07:36,220 --> 01:07:40,300
I will bring in some more of
these kinds of curious effects.

1319
01:07:40,300 --> 01:07:42,980
And then here's another one.

1320
01:07:42,980 --> 01:07:47,250
And it's very hard for you to
tell are the stairs going up?

1321
01:07:47,250 --> 01:07:48,560
Are they going down?

1322
01:07:48,560 --> 01:07:49,730
What's going on?

1323
01:07:49,730 --> 01:07:56,590
It's the same kind of
play with the paintings

1324
01:07:56,590 --> 01:07:59,990
to create confusion
in your perceptions.

1325
01:07:59,990 --> 01:08:02,060
And then here is
yet another one,

1326
01:08:02,060 --> 01:08:04,390
where you don't know
is water running up,

1327
01:08:04,390 --> 01:08:06,280
or is the water
running down here?

1328
01:08:06,280 --> 01:08:09,020
That's again an
interesting confusion

1329
01:08:09,020 --> 01:08:11,360
that Escher has created.

1330
01:08:11,360 --> 01:08:15,150
All right, so that
then is the essence

1331
01:08:15,150 --> 01:08:20,200
of what I wanted to
cover today to highlight

1332
01:08:20,200 --> 01:08:25,380
the fact that our
ability to extract shape

1333
01:08:25,380 --> 01:08:30,410
information in the world
is absolutely incredible.

1334
01:08:30,410 --> 01:08:33,479
And it has triggered
not only experiments

1335
01:08:33,479 --> 01:08:36,640
to try to understand
that by scientists,

1336
01:08:36,640 --> 01:08:41,810
but it has created a
tremendous amount of artwork

1337
01:08:41,810 --> 01:08:46,979
that played with these
because it was so enjoyable.

1338
01:08:46,979 --> 01:08:49,649
And certainly, the several
paintings that I have shown you

1339
01:08:49,649 --> 01:08:51,630
must have given you
a pretty good sense

1340
01:08:51,630 --> 01:08:57,660
of how artists capitalize on
these limitations in ability

1341
01:08:57,660 --> 01:09:02,550
to extract intermediate
and high level vision

1342
01:09:02,550 --> 01:09:05,490
capacities from
the visual scene.

1343
01:09:05,490 --> 01:09:08,830
All right, so to summarize
them, first of all,

1344
01:09:08,830 --> 01:09:12,470
I mentioned three
major theories that

1345
01:09:12,470 --> 01:09:14,350
have to do with form processing.

1346
01:09:14,350 --> 01:09:17,160
The one, the orientation
of line segments.

1347
01:09:17,160 --> 01:09:19,840
And I pointed out to you
that that theory is not

1348
01:09:19,840 --> 01:09:21,170
particularly powerful.

1349
01:09:21,170 --> 01:09:24,979
And then the one, the
topographic theory,

1350
01:09:24,979 --> 01:09:28,359
which turned out to be
bordering on the ridiculous.

1351
01:09:28,359 --> 01:09:30,350
And finally, Fourier
analysis, which

1352
01:09:30,350 --> 01:09:33,229
seems to have a lot of
power, even though it also

1353
01:09:33,229 --> 01:09:35,740
has created a lot of skeptics.

1354
01:09:35,740 --> 01:09:40,060
All right, then I pointed
out to that are the V2, V4,

1355
01:09:40,060 --> 01:09:43,779
and inferotemporal cortex,
play important roles

1356
01:09:43,779 --> 01:09:49,460
in intermediate vision, these
intermediate visual capacities

1357
01:09:49,460 --> 01:09:51,640
we talked about in detail.

1358
01:09:51,640 --> 01:09:57,020
Then I pointed out to
you also that in V2, it

1359
01:09:57,020 --> 01:09:59,900
was discovered that there
are some neurons that

1360
01:09:59,900 --> 01:10:06,170
respond to subjective contours,
indicating that already in area

1361
01:10:06,170 --> 01:10:09,470
V2, we can perform these
incredibly higher level

1362
01:10:09,470 --> 01:10:13,260
abilities to extract
information when

1363
01:10:13,260 --> 01:10:15,850
it is unclear in
the visual scene.

1364
01:10:15,850 --> 01:10:19,840
Then I pointed out to
you that recognition

1365
01:10:19,840 --> 01:10:22,390
of objects transformed
in the various ways

1366
01:10:22,390 --> 01:10:25,760
is compromised by V4 and
inferotemporal lesions.

1367
01:10:25,760 --> 01:10:28,970
And V4 lesions also
produce major deficits

1368
01:10:28,970 --> 01:10:34,100
in learning and selecting
lesser stimuli, which

1369
01:10:34,100 --> 01:10:35,980
is a very important
attribute for us

1370
01:10:35,980 --> 01:10:39,830
to be able not the respond
in reflex-like fashion

1371
01:10:39,830 --> 01:10:41,490
to what is most
obvious out there,

1372
01:10:41,490 --> 01:10:44,410
but to be able to
extract the subtle things

1373
01:10:44,410 --> 01:10:46,300
from the visual scene.

1374
01:10:46,300 --> 01:10:48,880
Some inferotemporal neurons
are selective for objects

1375
01:10:48,880 --> 01:10:50,140
including faces.

1376
01:10:50,140 --> 01:10:53,430
But most respond to a variety
of objects whose recognition is

1377
01:10:53,430 --> 01:10:55,920
based on the
differential activity

1378
01:10:55,920 --> 01:10:57,740
of a great many neurons.

1379
01:10:57,740 --> 01:10:58,820

1380
01:10:58,820 --> 01:11:03,860
OK, so that then brings me
to the last point, which

1381
01:11:03,860 --> 01:11:08,670
is how we process and deal
with ambiguities in perception,

1382
01:11:08,670 --> 01:11:11,850
unfortunately still
remains a mystery.

1383
01:11:11,850 --> 01:11:19,600
And so there's a lot of space
here for new investigators

1384
01:11:19,600 --> 01:11:24,170
to come up with exciting,
interesting new findings

1385
01:11:24,170 --> 01:11:31,440
about how the brain performs
these kinds of subtle analyses

1386
01:11:31,440 --> 01:11:32,657
of the visual scene.

1387
01:11:32,657 --> 01:11:33,820

1388
01:11:33,820 --> 01:11:38,250
OK, I think I'll leave
this until next time.

1389
01:11:38,250 --> 01:11:39,635
I'll talk about this next time.

1390
01:11:39,635 --> 01:11:41,380

1391
01:11:41,380 --> 01:11:43,891
OK, does anybody
have any questions?

1392
01:11:43,891 --> 01:11:47,820

1393
01:11:47,820 --> 01:11:50,400
All right, I hope that you
can take a little bit of time

1394
01:11:50,400 --> 01:11:53,540
out, look at some of these
artists on the internet,

1395
01:11:53,540 --> 01:12:00,410
and look at Hirschfeld and the
Ninas that I had shown you.

1396
01:12:00,410 --> 01:12:40,056