1
00:00:00,000 --> 00:00:06,255
[MUSIC]

2
00:00:06,255 --> 00:00:07,970
Hello, my name is Manuel Spectra.

3
00:00:07,970 --> 00:00:09,300
I work at the University of Oxford.

4
00:00:09,300 --> 00:00:12,220
And here I'm sitting with three experts in the field of color vision and

5
00:00:12,220 --> 00:00:13,450
color perception.

6
00:00:13,450 --> 00:00:17,950
Professor Professor Anne Hubbard and Professor David Brainerd.

7
00:00:17,950 --> 00:00:21,890
About three years ago, three and a half years ago, an image of the Internet.

8
00:00:21,890 --> 00:00:23,420
It was a viral image.

9
00:00:23,420 --> 00:00:27,970
It was a picture, a photograph of a dress, and it was quite polarizing.

10
00:00:27,970 --> 00:00:30,130
Anya, do you want to tell us a little bit about the dress?

11
00:00:30,130 --> 00:00:33,000
>> Yes, this dress divided people into two groups.

12
00:00:33,000 --> 00:00:36,300
Those who saw the dress in the photgraph as white and gold, and

13
00:00:36,300 --> 00:00:40,290
those who saw it as blue and black and huge arguments over this.

14
00:00:40,290 --> 00:00:42,430
People claim to be losing friends.

15
00:00:42,430 --> 00:00:44,500
Over this phenomenon.

16
00:00:44,500 --> 00:00:49,450
How could different people look at the same image of the same object and

17
00:00:49,450 --> 00:00:52,310
see it as these radically different colors.

18
00:00:52,310 --> 00:00:55,800
That's a question that excited the vision science community immediately.

19
00:00:55,800 --> 00:00:58,790
One thing we were so thrilled that everyone else suddenly became fascinated

20
00:00:58,790 --> 00:01:04,620
by this phenomenon of color perception and by the striking individual differences in

21
00:01:04,620 --> 00:01:10,060
color perception, which we think reveals the inherent subjectivity of color,

22
00:01:10,060 --> 00:01:14,170
as well as with further study some of the basic mechanisms that the visual system

23
00:01:14,170 --> 00:01:19,540
uses to make color a meaningful aspect of our visual perception.

24
00:01:19,540 --> 00:01:21,150
So one of the most exciting things for

25
00:01:21,150 --> 00:01:24,590
us as well about this photograph is that although people

26
00:01:24,590 --> 00:01:28,470
looked at the photograph and saw the dresses, different colors in the image.

27
00:01:28,470 --> 00:01:33,240
It's also the case that in the real world, the real original dress,

28
00:01:33,240 --> 00:01:37,630
put under slightly ambiguous lighting can evoke the same controversy.

29
00:01:37,630 --> 00:01:40,680
People looking at this same object but

30
00:01:40,680 --> 00:01:45,480
under ambiguous lighting will perceive it as being different colors.

31
00:01:45,480 --> 00:01:48,200
>> But I think it's worth saying that, if you were to put the dress on and

32
00:01:48,200 --> 00:01:52,600
just walk around, day to day, everybody would agree it was a blue and black dress.

33
00:01:52,600 --> 00:01:58,771
You have to go to some work [CROSSTALK] And fuss around to create this effect.

34
00:01:58,771 --> 00:02:02,390
And that actually, some ways tells us how well vision works most of the time.

35
00:02:02,390 --> 00:02:03,170
>> Exactly.

36
00:02:03,170 --> 00:02:04,470
>> Which is the thing that made it so

37
00:02:04,470 --> 00:02:08,790
surprising when people were seeing the image or the real dress.

38
00:02:08,790 --> 00:02:11,270
>> It's so interesting people say what is it really?

39
00:02:11,270 --> 00:02:14,640
But the question they're really asking is what does it really look like

40
00:02:14,640 --> 00:02:18,430
under natural broadband roughly white light?

41
00:02:18,430 --> 00:02:20,720
Because it's not what it is really.

42
00:02:20,720 --> 00:02:25,690
It has reflectants properties, which can reflect different light depending

43
00:02:25,690 --> 00:02:28,890
on the light that's shining on it and the context in which it is seen.

44
00:02:28,890 --> 00:02:32,330
And the visuals has been constantly trying to resolve that ambiguity

45
00:02:32,330 --> 00:02:32,914
between >> And

46
00:02:32,914 --> 00:02:39,160
it usually does it in an almost perfect way where basically we're happy to call

47
00:02:39,160 --> 00:02:44,430
objects of different color names like a blue top or a brown jacket.

48
00:02:44,430 --> 00:02:47,760
>> Brown, that's great.

49
00:02:47,760 --> 00:02:48,725
>> No, it's not great.

50
00:02:48,725 --> 00:02:51,060
>> [LAUGH] >> Exactly right.

51
00:02:51,060 --> 00:02:53,670
>> Yeah, there are minor differences.

52
00:02:53,670 --> 00:02:56,210
But overall, we doing a really good, great job and

53
00:02:56,210 --> 00:03:01,900
we're not aware that the color isn't really a property of the object.

54
00:03:01,900 --> 00:03:05,020
Because the light that gets into the eyes is the combination

55
00:03:05,020 --> 00:03:08,360
of the reflecting properties of the objects and the illumination and

56
00:03:08,360 --> 00:03:10,860
the illumination changes throughout the day.

57
00:03:10,860 --> 00:03:12,660
And as we enter the room.

58
00:03:12,660 --> 00:03:16,880
>> What you can see here from this little demo is that the light

59
00:03:16,880 --> 00:03:20,670
on the dress constantly changing in color is causing the light reflected from

60
00:03:20,670 --> 00:03:22,430
the dress to constantly change.

61
00:03:22,430 --> 00:03:25,480
>> So you mentioned color constancy or constancy here.

62
00:03:25,480 --> 00:03:28,330
Can you unpack what that term means and

63
00:03:28,330 --> 00:03:32,230
how important it is in basically visual processing?

64
00:03:32,230 --> 00:03:36,310
>> Well, we think of vision is serving a purpose and telling us what is where.

65
00:03:36,310 --> 00:03:37,580
What's out there.

66
00:03:37,580 --> 00:03:39,560
And where is it so that we can interact with it?

67
00:03:39,560 --> 00:03:44,930
And one of the things color constancy does is enable us to use color as a meaningful,

68
00:03:44,930 --> 00:03:47,030
reliable cue to what an object is.

69
00:03:47,030 --> 00:03:51,710
So the colors we perceive objects to be correspond pretty

70
00:03:51,710 --> 00:03:56,220
invariably to their actual reflectance properties, which tell us something about

71
00:03:56,220 --> 00:03:59,360
their material properties, and therefore tells us something about what they are.

72
00:03:59,360 --> 00:04:04,180
But that set of processes is actually very complex and happens on multiple levels.

73
00:04:04,180 --> 00:04:06,080
We're just unconscious of it.

74
00:04:06,080 --> 00:04:10,060
We need to get rid of the contaminating effects of the illumination spectrum

75
00:04:10,060 --> 00:04:14,610
effectively, and get something closer to the actual reflecting

76
00:04:14,610 --> 00:04:16,310
physical properties of the object.

77
00:04:16,310 --> 00:04:18,420
>> But that's not a stimulus that we ever see, right?

78
00:04:18,420 --> 00:04:21,990
We never perceive the reflectants of an object, right?

79
00:04:21,990 --> 00:04:25,810
So it's not in a sense contaminated signal,[CROSS TALK]

80
00:04:25,810 --> 00:04:26,750
it is the only signal that we have.

81
00:04:26,750 --> 00:04:30,790
>> The true is that the reflectance the dress itself doesn't jump into our eyes,

82
00:04:30,790 --> 00:04:33,480
not the physical object interacts with the retina.

83
00:04:33,480 --> 00:04:36,060
It's the light reflected from the object.

84
00:04:36,060 --> 00:04:39,260
And that light brings the information about the properties of the object

85
00:04:39,260 --> 00:04:42,570
the dress, its color, its size and shape its texture.

86
00:04:42,570 --> 00:04:46,420
To our eyes, but it brings it in a form that contaminates, as I may say,

87
00:04:46,420 --> 00:04:49,530
that's contaminated or intermixed with other things, like what

88
00:04:49,530 --> 00:04:53,470
is the spectrum of the illumination and how is it rotated with respect to us.

89
00:04:53,470 --> 00:04:56,850
And I'd say it's worth saying, and this is mostly about color, but

90
00:04:56,850 --> 00:05:01,850
it's worth saying that these issues arise with all aspects of perception the size

91
00:05:01,850 --> 00:05:05,500
of an image on our retina changes with its distance, and yet we don't see

92
00:05:05,500 --> 00:05:08,710
the dress growing and shrinking as we move closer or further from it.

93
00:05:08,710 --> 00:05:11,150
And all that's the work of our brain

94
00:05:11,150 --> 00:05:14,950
delivering to us percepts that are useful about the world.

95
00:05:14,950 --> 00:05:18,320
>> Vision isn't passive process.

96
00:05:18,320 --> 00:05:20,710
We don't just see what's out there.

97
00:05:20,710 --> 00:05:25,400
Were constantly actively constructing hypothesis about what's out there

98
00:05:25,400 --> 00:05:30,610
from the inherently ambiguous light signals that the eye receives.

99
00:05:30,610 --> 00:05:33,350
>> So what specifically was ambiguous then

100
00:05:33,350 --> 00:05:35,906
in particular about the photograph of the dress?

101
00:05:35,906 --> 00:05:41,470
>> Well, it wasn't obvious because there weren't very many

102
00:05:41,470 --> 00:05:46,610
cues the photograph as to what the color of the light source was on the dress,

103
00:05:46,610 --> 00:05:49,850
and it's highly likely that there were actually two different light sources on

104
00:05:49,850 --> 00:05:55,650
the dress, a bluish daylight and a yellowish, probably artificial light.

105
00:05:55,650 --> 00:06:01,020
And this meant that when people's inherent color constancy mechanisms

106
00:06:01,020 --> 00:06:04,900
effectively corrected for the color of the light, that could go one way or the other.

107
00:06:04,900 --> 00:06:08,570
And in fact, subsequent experiments have shown that people who

108
00:06:08,570 --> 00:06:12,420
effectively assumed that the light shining on the dress was a sort of dim

109
00:06:12,420 --> 00:06:17,210
blue light saw the dresses white and gold, and those people who effectively assumed

110
00:06:17,210 --> 00:06:20,130
unconsciously that the light shining and the dress was a bright yellow.

111
00:06:20,130 --> 00:06:22,230
They reported it to be blue and black.

112
00:06:22,230 --> 00:06:26,820
So those perceptions that people reported actually give us

113
00:06:26,820 --> 00:06:31,210
access into the internal workings of their color constancy mechanisms.

114
00:06:31,210 --> 00:06:35,910
>> But you know, I would say that, I think I agree with your conclusion, but

115
00:06:35,910 --> 00:06:40,570
I'll say that, when this phenomenon first came about,

116
00:06:40,570 --> 00:06:43,380
the thought process, at least that I started to go through was

117
00:06:43,380 --> 00:06:46,120
not immediately to jump to that conclusion, but rather to ask,

118
00:06:46,120 --> 00:06:48,920
well here is a case where people are reporting different colors.

119
00:06:48,920 --> 00:06:51,520
What is it we know about color vision?

120
00:06:51,520 --> 00:06:54,430
And the way that light is encoded might or might not explain that.

121
00:06:54,430 --> 00:06:55,500
What would you wanna rule out?

122
00:06:55,500 --> 00:06:59,420
Because most of the time we go through lights agreeing about colors and

123
00:06:59,420 --> 00:07:01,140
something was different here.

124
00:07:01,140 --> 00:07:04,950
And so I think it's worth talking a little bit about what we know about

125
00:07:04,950 --> 00:07:07,610
the encoding of light by the retina.

126
00:07:07,610 --> 00:07:09,420
Is that the same across people?

127
00:07:09,420 --> 00:07:14,940
And what other factors one ruled out before one sort of got this idea about

128
00:07:14,940 --> 00:07:19,060
color constancy in the dress, which I think is now the best idea we have.

129
00:07:19,060 --> 00:07:22,640
>> One of the very first things was of course, the idea that people would look at

130
00:07:22,640 --> 00:07:25,395
this image of the dress, just on different display devices.

131
00:07:25,395 --> 00:07:25,990
>> Mm-hm. >> Yeah.

132
00:07:25,990 --> 00:07:28,930
>> Which would differently calibrate it, and that would lead to differences.

133
00:07:28,930 --> 00:07:29,820
But that could mean extruding it.

134
00:07:29,820 --> 00:07:31,060
>> I remember clicking my phone and rotating it.

135
00:07:31,060 --> 00:07:32,190
>> Yes, yes. >> You could change it

136
00:07:32,190 --> 00:07:33,390
from blue/black to white/gold.

137
00:07:33,390 --> 00:07:33,900
>> Yeah.

138
00:07:33,900 --> 00:07:35,800
>> Because the display had an angular effect.

139
00:07:35,800 --> 00:07:37,930
>> Yeah, and there was some notion, as well,

140
00:07:37,930 --> 00:07:40,020
that it might be a difference between old people and

141
00:07:40,020 --> 00:07:44,300
young people, people who had a bit more, maybe cataract, than young people.

142
00:07:44,300 --> 00:07:45,260
>> So it was- >> Or a difference of

143
00:07:45,260 --> 00:07:47,520
viewing it in artificial light versus daylight.

144
00:07:47,520 --> 00:07:50,620
>> So one of the first thing that happened actually were reports on the internet that

145
00:07:50,620 --> 00:07:55,050
somebody in vision science put one up on a projector and got 15 people to come in

146
00:07:55,050 --> 00:07:58,700
look at the same image and the same display from the same angle.

147
00:07:58,700 --> 00:08:00,355
And they still were fighting with each other about it.

148
00:08:00,355 --> 00:08:02,400
>> [LAUGH] >> So, you know, that was, you know,

149
00:08:02,400 --> 00:08:05,930
good tool out a potential just display artifact.

150
00:08:05,930 --> 00:08:07,980
>> Or was it just a difference in naming?

151
00:08:07,980 --> 00:08:10,740
Everybody was actually seeing the same thing, but some people

152
00:08:10,740 --> 00:08:15,550
call bluish whites blue and other people call bluish whites white perhaps.

153
00:08:15,550 --> 00:08:19,990
>> So that's when we got a couple of observers into the lab and

154
00:08:19,990 --> 00:08:25,210
actually had them do color matches to the dress under controlled conditions, and

155
00:08:25,210 --> 00:08:27,306
we still found a difference.

156
00:08:27,306 --> 00:08:30,750
So that's a good,

157
00:08:30,750 --> 00:08:34,970
that supports the idea that they really perceived the dress to be different.

158
00:08:34,970 --> 00:08:37,711
>> It wasn't just a difference in the way they choose to name something that they

159
00:08:37,711 --> 00:08:38,377
[CROSSTALK] >> And

160
00:08:38,377 --> 00:08:42,202
actually it would've been really surprising.

161
00:08:42,202 --> 00:08:45,090
That report was I remember one of the earlier reports and very important.

162
00:08:45,090 --> 00:08:48,530
But it would've been really surprising if it had been different just because

163
00:08:48,530 --> 00:08:52,430
there have been lots of studies of color naming of little well controled chips and

164
00:08:52,430 --> 00:08:53,700
isolation.

165
00:08:53,700 --> 00:08:54,270
And you know,

166
00:08:54,270 --> 00:08:57,900
the whole paint industry relies on the fact that we can look at paint chips and

167
00:08:57,900 --> 00:09:02,350
agree about their names and develop a nomenclature that people agree on.

168
00:09:02,350 --> 00:09:05,985
So, it had just been naming your it would have been a mystery

169
00:09:05,985 --> 00:09:09,915
as to why suddenly with the dress where the names going differently.

170
00:09:09,915 --> 00:09:12,775
>> But there are cases in which this is the same color will be named

171
00:09:12,775 --> 00:09:15,033
in different ways by different people right?

172
00:09:15,033 --> 00:09:16,155
>> Subtly, yeah.

173
00:09:16,155 --> 00:09:17,105
>> Differences for

174
00:09:17,105 --> 00:09:20,775
certain color combinations like the blue green turquoise.

175
00:09:20,775 --> 00:09:24,670
Some people use the turquoise category as a stoned or between the.

176
00:09:24,670 --> 00:09:26,370
Blue and purple there.

177
00:09:26,370 --> 00:09:29,100
There are different light differences between people for

178
00:09:29,100 --> 00:09:32,600
all the other categories I've seen, they agree really quite a lot.

179
00:09:32,600 --> 00:09:36,540
>> Well the interesting thing is that blue is actually a huge category so especially

180
00:09:36,540 --> 00:09:41,010
in Western cultures, light blues, mid blues and dark blues all coalesce into one

181
00:09:41,010 --> 00:09:46,040
big, big category whereas in other cultures for example, in Russian, they're

182
00:09:46,040 --> 00:09:51,780
divided into two very distinct categories with different color terms for them.

183
00:09:51,780 --> 00:09:55,520
But those other sort of smaller categories, the reds, oranges, browns,

184
00:09:55,520 --> 00:09:59,380
yellows, they tend to be more consistently categorized.

185
00:09:59,380 --> 00:10:02,110
>> And also those are colors that when make people make some way

186
00:10:02,110 --> 00:10:02,710
already judgments.

187
00:10:02,710 --> 00:10:06,220
It's pushing a boundary around blue, green.

188
00:10:06,220 --> 00:10:09,720
Everybody agrees that blue is similar to green, someone's a little blue or

189
00:10:09,720 --> 00:10:11,430
a little greener in their naming.

190
00:10:11,430 --> 00:10:14,080
It isn't the cause of fistfights.

191
00:10:14,080 --> 00:10:17,900
The way, that the way that the it's not true of the disagreements about the dress

192
00:10:17,900 --> 00:10:20,110
color, they weren't just.

193
00:10:20,110 --> 00:10:24,860
The boundary of colors that were similar to begin with.

194
00:10:24,860 --> 00:10:28,030
>> White and gold sounds radically different from blue and black, and

195
00:10:28,030 --> 00:10:30,920
that was one of the reasons why that caused such conflict.

196
00:10:30,920 --> 00:10:35,850
If you really dig into it you find that actually whitish blues are difficult

197
00:10:35,850 --> 00:10:39,410
to classify and a lot of people do put them into different categories,

198
00:10:39,410 --> 00:10:43,500
and blue-ish white can be called a blue or white.

199
00:10:43,500 --> 00:10:46,320
And in fact in Western cultures,

200
00:10:46,320 --> 00:10:51,010
most people tend to think of their whites as being slightly bluish.

201
00:10:51,010 --> 00:10:52,860
It's an interesting- >> But golds and blacks?

202
00:10:52,860 --> 00:10:54,480
>> Golds and blacks, and that's another very interesting.

203
00:10:54,480 --> 00:10:58,930
And again, if you look at the dress, the lace, the black lace is quite shiny.

204
00:10:58,930 --> 00:11:01,940
When a yellowish light shines on it, you get yellow highlights.

205
00:11:01,940 --> 00:11:05,300
That sort of dark color can start to look like a sort of dark gold.

206
00:11:05,300 --> 00:11:07,850
>> Mm hm >> So it's actually not that far away,

207
00:11:07,850 --> 00:11:08,990
a gold and a black.

208
00:11:08,990 --> 00:11:15,400
But that's digging into it and getting into very sort of knit picky details it.

209
00:11:15,400 --> 00:11:19,220
The fundamental phenomenon is that color perception is subjective but

210
00:11:19,220 --> 00:11:21,200
relates to objective physical properties.

211
00:11:21,200 --> 00:11:23,900
>> So there's an interesting aspect here which is that sort of obviously you can

212
00:11:23,900 --> 00:11:28,490
have, say a surface that's painted black or

213
00:11:28,490 --> 00:11:30,920
a matte wall is painted blue in some way.

214
00:11:30,920 --> 00:11:32,040
But gold,

215
00:11:32,040 --> 00:11:36,590
the use of the term gold points to another aspect I think of the dress which is,

216
00:11:36,590 --> 00:11:41,220
it's not just about reflectance is about more complex material properties as well.

217
00:11:41,220 --> 00:11:44,290
So where, in the field of vision science clearly,

218
00:11:44,290 --> 00:11:49,500
we have ways to study material perception as well.

219
00:11:49,500 --> 00:11:54,680
And in a sense, a lot of the work that has been done on

220
00:11:54,680 --> 00:11:58,170
color vision is relevant for material perception as well.

221
00:11:58,170 --> 00:12:03,550
So I was wondering if there's sort of, where do we stand with thinking about

222
00:12:03,550 --> 00:12:07,650
how we perceive complex materials in the world?

223
00:12:07,650 --> 00:12:12,580
>> Well, the color is just one aspect of the material properties of objects and

224
00:12:12,580 --> 00:12:15,380
we picked that up from the reflected light signal, but

225
00:12:15,380 --> 00:12:19,480
we can pick up many other properties about objects from that reflected light signal.

226
00:12:19,480 --> 00:12:25,870
The way it changes with viewpoint tells us whether that surface is glossy or mat.

227
00:12:25,870 --> 00:12:29,090
The way it changes with respect to other objects in the scene tells us

228
00:12:29,090 --> 00:12:33,610
whether the object is transparent or translucent or opaque.

229
00:12:33,610 --> 00:12:39,150
We have luster as well we have metallic versus non metallic reflection.

230
00:12:39,150 --> 00:12:43,800
All of these things can be gleaned from the light signal and

231
00:12:43,800 --> 00:12:48,860
the actual body reflectance of the object which is what we associate with its color,

232
00:12:48,860 --> 00:12:52,590
is just one of those physical properties, >> You know but, go ahead.

233
00:12:52,590 --> 00:12:56,070
>> Perception of material properties is really an emerging research field and

234
00:12:56,070 --> 00:12:57,130
during the last decade.

235
00:12:57,130 --> 00:13:02,150
I mean we've learned that our visual system is capable of extracting enormous

236
00:13:02,150 --> 00:13:07,600
lot of different properties from the input signal like whether surface is smooth,

237
00:13:07,600 --> 00:13:14,880
or rough, or whether it's hard or even whether it's hard or soft.

238
00:13:14,880 --> 00:13:20,370
I mean we learn these associations and we use certain

239
00:13:20,370 --> 00:13:25,250
properties of the inputs even certain queues that we don't even know yet

240
00:13:25,250 --> 00:13:29,740
we're in the process of finding out which ones they are to extract these

241
00:13:29,740 --> 00:13:33,120
very complex properties of the world.

242
00:13:33,120 --> 00:13:34,580
>> I was gonna go the other way though.

243
00:13:34,580 --> 00:13:39,000
Talk about not about not about high level what I would call higher and

244
00:13:39,000 --> 00:13:41,460
mid level visual aspects and complexities, but

245
00:13:41,460 --> 00:13:44,330
actually about the basic encoding of light and something that I think

246
00:13:44,330 --> 00:13:49,210
makes the color differences the dress surprising is we know among most people,

247
00:13:49,210 --> 00:13:54,140
the actual cells that encode the light and MCI this first step of the signal that

248
00:13:54,140 --> 00:13:58,890
gets to us are actually in common among most people so called color normal.

249
00:13:58,890 --> 00:14:05,390
Vision which has maybe 95, maybe more than that 98% of us.

250
00:14:05,390 --> 00:14:08,886
And the fact that that encoding is highly similar means that things like

251
00:14:08,886 --> 00:14:12,265
the way that the colors get mixed on your TV's like the red, green and

252
00:14:12,265 --> 00:14:15,190
blue primaries to make yellow is something we all agree on.

253
00:14:15,190 --> 00:14:17,660
You don't have to have your television customized.

254
00:14:17,660 --> 00:14:20,860
For each observer, because the way their color mixing behavior and

255
00:14:20,860 --> 00:14:22,680
perception works differs.

256
00:14:22,680 --> 00:14:26,450
And so that's one of the things I think, again, that suggests the commonality to

257
00:14:26,450 --> 00:14:30,020
color perception that makes the dress quite surprising.

258
00:14:30,020 --> 00:14:35,020
And again, as that phenomenon first came out I think people scratched

259
00:14:35,020 --> 00:14:40,330
their head and said, well could it be that maybe this particular stimulus

260
00:14:40,330 --> 00:14:43,070
is just on the edge of some normal cover difference.

261
00:14:43,070 --> 00:14:45,830
But I think the evidence now was accumulated from larger studies that

262
00:14:45,830 --> 00:14:47,570
that's not what's going on.

263
00:14:47,570 --> 00:14:52,630
But again, one could imagine those kinds of things and I think many of us did.

264
00:14:52,630 --> 00:14:56,560
>> Yes, so the dress isn't due to some differences in the eye.

265
00:14:56,560 --> 00:14:57,320
>> Right.

266
00:14:57,320 --> 00:14:59,540
>> In the actual receptors, the- >> Correct, yeah.

267
00:14:59,540 --> 00:15:02,990
>> The light sensors that are receiving the light signal.

268
00:15:02,990 --> 00:15:05,630
Or in, for example, the optics of the eye,

269
00:15:05,630 --> 00:15:08,770
in a certain cloudiness of the lens in some people versus other people.

270
00:15:08,770 --> 00:15:09,850
>> Right. >> Completely agree with you,

271
00:15:09,850 --> 00:15:11,910
very important to rule those things out.

272
00:15:11,910 --> 00:15:17,010
But then of course you do have the roughly 8% of males who

273
00:15:17,010 --> 00:15:19,300
do have differences in the light senses.

274
00:15:19,300 --> 00:15:19,871
>> Right. >> In their eye.

275
00:15:19,871 --> 00:15:22,100
>> Right. >> People with color vision deficiency.

276
00:15:22,100 --> 00:15:28,280
And that's another really hugely important source of individual differences which

277
00:15:28,280 --> 00:15:33,860
I think possibly doesn't receive enough attention in society.

278
00:15:33,860 --> 00:15:38,910
And should be receiving even more now that we use colors so

279
00:15:38,910 --> 00:15:42,900
much more for all of our various displace and

280
00:15:42,900 --> 00:15:48,990
devices that colors becoming ever more rampant, I could say.

281
00:15:48,990 --> 00:15:51,100
They say. [LAUGH]

282
00:15:51,100 --> 00:15:55,247
>> Stop, stop because they're small.

283
00:15:55,247 --> 00:15:59,119
>> I mean, the democratic influence basically have the problem that we had

284
00:15:59,119 --> 00:16:00,650
with the dress all the time.

285
00:16:00,650 --> 00:16:04,000
>> Yes. >> I mean, just like a whole dimension of

286
00:16:04,000 --> 00:16:06,620
the color representation.

287
00:16:06,620 --> 00:16:08,670
And when we talk about red or

288
00:16:08,670 --> 00:16:12,250
green, they don't know what we're talking about most of the time.

289
00:16:12,250 --> 00:16:15,800
So, but somehow they seem to cope.

290
00:16:15,800 --> 00:16:17,180
>> Well that's the astonishing thing.

291
00:16:17,180 --> 00:16:21,300
I think that there is an advantage of green rather.

292
00:16:21,300 --> 00:16:25,890
Crotchety old ophthalmologist in the early 20th century said that you had to really

293
00:16:25,890 --> 00:16:28,410
watch colorblind observers because they were very

294
00:16:28,410 --> 00:16:29,580
devious and- >> [LAUGH]

295
00:16:29,580 --> 00:16:31,230
>> He said they were basically-

296
00:16:31,230 --> 00:16:32,250
>> Trained from a young age.

297
00:16:32,250 --> 00:16:35,490
>> Yeah, trained from a young age to learn how to give the right answer and

298
00:16:35,490 --> 00:16:39,230
they were actually super intelligent because they go through life.

299
00:16:39,230 --> 00:16:41,710
Seeing the world completely differently from other people, but

300
00:16:41,710 --> 00:16:43,410
somehow keep this hidden.

301
00:16:43,410 --> 00:16:45,560
I think it was possibly easier then.

302
00:16:45,560 --> 00:16:49,320
And I am concerned now that kids in schools who

303
00:16:49,320 --> 00:16:55,370
are exposed to increasing amounts of color coding will be,

304
00:16:55,370 --> 00:16:58,540
in some ways more disadvantaged by their color vision deficiency,

305
00:16:58,540 --> 00:17:04,090
their inability to discriminate reds, from greens and browns from khakis.

306
00:17:04,090 --> 00:17:08,160
>> But so in the past apart from this idea that there's people who will say lack one

307
00:17:08,160 --> 00:17:12,880
specific photo pigment, there's also variation, natural variation or

308
00:17:12,880 --> 00:17:16,970
individual differences between people who would normally be classified as color

309
00:17:16,970 --> 00:17:18,560
normal observers.

310
00:17:18,560 --> 00:17:21,430
Yet at the same time, we they don't,

311
00:17:21,430 --> 00:17:28,160
I mean it's not as significant, perhaps, in terms of color reproduction.

312
00:17:28,160 --> 00:17:30,830
>> Well, these things, they come in degrees, and

313
00:17:30,830 --> 00:17:34,580
there's a small variation amongst when you look at the little plates and

314
00:17:34,580 --> 00:17:37,720
you read the number off or trace the shape.

315
00:17:37,720 --> 00:17:40,810
That classifies people as color normal, and there's a small variation there.

316
00:17:40,810 --> 00:17:44,670
But there is also maybe this is what you're referring to is cover anomalous

317
00:17:44,670 --> 00:17:47,510
individuals who are not colorblind,

318
00:17:47,510 --> 00:17:52,020
we might say their color different, and whose precepts again won't match

319
00:17:52,020 --> 00:17:56,540
the majority of the populations in the 8% of mainly males that I mentioned.

320
00:17:56,540 --> 00:18:02,630
So that's another source of individual variation, I think is well understood,

321
00:18:02,630 --> 00:18:06,590
not related to color consistency directly [be]cause it's at the visual front end.

322
00:18:06,590 --> 00:18:09,450
>> I think it's quite amazing how the visual system calibrates

323
00:18:09,450 --> 00:18:11,750
itself into the world.

324
00:18:11,750 --> 00:18:16,740
There's enormous variation in the relative number of L and M cones, or red and

325
00:18:16,740 --> 00:18:19,770
green cones that we have in the retina.

326
00:18:19,770 --> 00:18:24,380
And still, when people are asked, as David has shown us very nicely, to adjust

327
00:18:24,380 --> 00:18:28,870
a unit yellow, to mix a yellow out of red and green, they all agree perfectly well.

328
00:18:28,870 --> 00:18:31,130
And I think that's truly amazing.

329
00:18:31,130 --> 00:18:32,190
>> It is astonishing, and

330
00:18:32,190 --> 00:18:36,240
I think we still don't really understand how that comes about.

331
00:18:36,240 --> 00:18:38,830
How that calibration of a visual system comes about.

332
00:18:38,830 --> 00:18:41,900
I mean clearly there's an earlier experience, there's an environmental

333
00:18:41,900 --> 00:18:46,950
influence, and there's learning, but what is it that enables one

334
00:18:46,950 --> 00:18:51,750
person's unique yellow to match another person's unique yellow, despite really big

335
00:18:51,750 --> 00:18:56,150
differences in the way their light sensors are laid out in the back of the retina?

336
00:18:56,150 --> 00:18:56,880
I agree with you.

337
00:18:56,880 --> 00:19:00,980
I think that's a fascinating and unexplained phenomenon.

338
00:19:00,980 --> 00:19:05,310
>> I think it seems like there's a lot of research that could be done in that area,

339
00:19:05,310 --> 00:19:06,250
absolutely.

340
00:19:06,250 --> 00:19:09,748
Now, this sort of circles back to the dress, but there's the obvious question,

341
00:19:09,748 --> 00:19:13,400
well, to a color-deficient observer, have there been any studies looking at color

342
00:19:13,400 --> 00:19:18,560
naming of the dress in people's color vision deficiencies?

343
00:19:18,560 --> 00:19:20,705
>> Only anecdotally, to my knowledge.

344
00:19:20,705 --> 00:19:26,740
>> [LAUGH] >> Interestingly,

345
00:19:26,740 --> 00:19:32,340
some people with color vision deficiency distance themselves from the phenomenon,

346
00:19:32,340 --> 00:19:34,500
don't engage in the debate.

347
00:19:34,500 --> 00:19:36,020
>> They're already used to people telling them.

348
00:19:36,020 --> 00:19:38,375
>> They're so used to that already.

349
00:19:38,375 --> 00:19:42,100
>> [LAUGH] >> So I suppose one of the some

350
00:19:42,100 --> 00:19:46,340
interesting questions for me as an early career researcher in this field is sort of

351
00:19:46,340 --> 00:19:49,960
to hear from obviously more senior people such as you

352
00:19:49,960 --> 00:19:54,490
episode where you see the field going, seeing the next ten to 20 years, and

353
00:19:54,490 --> 00:19:59,260
what the big sort of emerging topics we heard material perception was one of them.

354
00:19:59,260 --> 00:20:00,590
But what's so

355
00:20:00,590 --> 00:20:06,930
where are we heading with color science or the neuroscience of color vision.

356
00:20:06,930 --> 00:20:08,820
>> I think it's heading into different directions.

357
00:20:08,820 --> 00:20:13,110
Of course, there is many more really interesting questions in the retina

358
00:20:13,110 --> 00:20:16,970
that people are starting to look into now using very complex and

359
00:20:16,970 --> 00:20:18,980
fancy new technology like adaptive optics.

360
00:20:18,980 --> 00:20:23,640
But it's also, we're more capable now of studying color vision in the real world or

361
00:20:23,640 --> 00:20:28,630
in virtual realities where we can vary the illuminant as we wish I

362
00:20:28,630 --> 00:20:32,520
think that opens up a lot of new possibilities.

363
00:20:32,520 --> 00:20:35,420
>> I want to second actually, both of the things you said but particularly,

364
00:20:35,420 --> 00:20:41,190
you know, something that emphasized by the phenomenal address is that in a complex

365
00:20:41,190 --> 00:20:44,690
image, I should maybe I should back up and say a lot of what we know about color

366
00:20:44,690 --> 00:20:50,300
vision is knowledge obtained where people looked at very well controlled flat,

367
00:20:50,300 --> 00:20:54,240
spatially uniform discs and maybe a background and if you were really doing

368
00:20:54,240 --> 00:20:57,690
a complicated experiment, the color of the background would change.

369
00:20:57,690 --> 00:20:59,390
That was a radical maneuver.

370
00:20:59,390 --> 00:21:05,510
You get to the dress and now it's a very spatially complex image of a real thing.

371
00:21:05,510 --> 00:21:11,180
With material and mixed with color with the color attached to an object and

372
00:21:11,180 --> 00:21:16,830
I'd say our understanding of how color maps out into this complex objects

373
00:21:16,830 --> 00:21:20,120
is in something Anya has emphasized is that

374
00:21:20,120 --> 00:21:24,330
we get most real objects even though we give them a code name they often,

375
00:21:24,330 --> 00:21:27,240
if you look at the individual little bits, there's lots of colors there that somehow,

376
00:21:27,240 --> 00:21:29,800
you grow them together in a way that's not fully understood.

377
00:21:29,800 --> 00:21:33,860
So that direction, I think, is an exciting avenue, one, I'd say,

378
00:21:33,860 --> 00:21:36,490
where quite what the right experiment to do and

379
00:21:36,490 --> 00:21:38,560
the right way to think about it is something we're struggling with.

380
00:21:38,560 --> 00:21:42,250
So there's a risk as the young researcher there's a risk that it might not work out.

381
00:21:42,250 --> 00:21:44,650
But the questions are really interesting.

382
00:21:44,650 --> 00:21:48,970
>> Yeah, yes, so I think to understand how the perception

383
00:21:48,970 --> 00:21:52,910
of object color is integrated with the perception of the object.

384
00:21:52,910 --> 00:21:58,720
And the scene in the world as a whole, in the natural world, in real,

385
00:21:58,720 --> 00:22:03,830
natural behavioral tasks is still a massive goal.

386
00:22:03,830 --> 00:22:09,990
And one which we do have better tools to tackle, including

387
00:22:09,990 --> 00:22:14,700
better computer graphics, so better control over computer generated stimuli,

388
00:22:14,700 --> 00:22:24,090
and better control over real-world stimuli, multi-spectrum light sources.

389
00:22:24,090 --> 00:22:27,400
Specialist imaging equipment.

390
00:22:27,400 --> 00:22:30,660
All these things give us better control over actual stimuli,

391
00:22:30,660 --> 00:22:35,526
which more closely replicate those natural stimuli with which we interact every day.

392
00:22:35,526 --> 00:22:41,020
But I think there's still debate

393
00:22:41,020 --> 00:22:46,755
over the actual function of color in real world object perception.

394
00:22:46,755 --> 00:22:51,740
>> Something I would say, coming back to the dress and

395
00:22:51,740 --> 00:22:55,700
these points is that, you know we have a challenge to with the dress which is there

396
00:22:55,700 --> 00:23:01,320
is this one image that has this striking behavior and people are there was a talk

397
00:23:01,320 --> 00:23:07,130
about the shore in VSS this year, which was another object that had this bicolor.

398
00:23:07,130 --> 00:23:09,880
And people are beginning to work on

399
00:23:09,880 --> 00:23:13,430
synthesizing stimuli that produce these fights.

400
00:23:13,430 --> 00:23:15,930
So you know, maybe making the world worse temporarily.

401
00:23:15,930 --> 00:23:20,510
But when we understand enough about it so that we can create the stimuli and

402
00:23:20,510 --> 00:23:23,160
manipulate, is it gonna be ambiguous or is it not?

403
00:23:23,160 --> 00:23:26,570
Then we really begin to understand what it is that's causing the ambiguity.

404
00:23:26,570 --> 00:23:28,540
At the moment, I think we're still searching for

405
00:23:28,540 --> 00:23:31,360
the detailed clues that will let us figure that out.

406
00:23:31,360 --> 00:23:36,830
>> And conversely again, going back to the dress if you take the real dress.

407
00:23:36,830 --> 00:23:39,560
And make different environmental manipulations and

408
00:23:39,560 --> 00:23:43,590
determine which of those influence people into seeing it the same way.

409
00:23:43,590 --> 00:23:47,770
You can get a better handle on what those key mechanisms are that contribute to

410
00:23:47,770 --> 00:23:49,070
color constancy.

411
00:23:49,070 --> 00:23:52,380
So for example, taking the real dress and putting in blue light and

412
00:23:52,380 --> 00:23:55,010
then yellow light on it, you will get people to disagree.

413
00:23:55,010 --> 00:23:57,520
But if you put it against a yellow background as in Dr.

414
00:23:57,520 --> 00:24:01,400
Verner has shown, you'll push them on the direction of seeing it as blue black.

415
00:24:01,400 --> 00:24:03,360
Or if you just light the edges with the blue light,

416
00:24:03,360 --> 00:24:05,510
you'll push them in the direction and see it in white and gold,

417
00:24:05,510 --> 00:24:09,030
so then we know that background is important and edges are important.

418
00:24:09,030 --> 00:24:14,003
And we can learn then, a bit more about those critical, real word cues

419
00:24:14,003 --> 00:24:18,998
that the human visual system uses in practicing its color constancy.

420
00:24:18,998 --> 00:24:24,528
>> So in summary in a sense it's sort of more experimental

421
00:24:24,528 --> 00:24:30,541
control we can really tease apart [INAUDIBLE] So thank you for

422
00:24:30,541 --> 00:24:36,770
joining me and this discussion was very interesting.

423
00:24:36,770 --> 00:24:48,640
And thanks for watching.