People are freaking out over what color the dress “really is”. They’re projecting the property of “true color” onto the real world, when in reality “true color” is only in their mind.
Human language isn’t great for talking about colors. The word “blue”, for example, encompasses very different hues.
Let’s think about this in terms of the HSV (hue, saturation, value) model.
In this model, whitish colors have low saturation, high value (value is basically brightness), and the hue doesn’t matter. Similarly, blackish colors have high saturation, low value, and the hue doesn’t matter again.
The dress in this image has the blue hue, but high value (=brightness). That makes it light blue or bluish white—take your pick. The other color has the orange hue, but low value. That makes it dark brown (gold) or brownish black, again, take your pick.
Overlay on top of this the differences in uncalibrated monitors, personal idiosyncrasies, and some optical illusion effects, and you have a ridiculously successfull buzzfeed post :-D
Yeah, that’s a pale blue, but I find it hard to alieve (and, had I not read Generalizing From One Example, I’d also find it hard to believe) that a sizeable fraction of the population find it pale enough to call it “white” non-trollingly. This also applies if I look at it on a dark background such as the left panel of XKCD with my screen’s luminosity set to the maximum. (The only condition in which I would call such a thing “white” is if I thought it only looked like that because of the ambient light, but under so blue a light the dark parts of the dress would look quite black with hardly any trace of yellow whatsoever.)
People are freaking out over what color the dress “really is”.
Are they, actually? Or are they freaking out over what other people say it is?
If I saw only the text of the supposed controversy I would instantly diagnose it as a hoax for clickbait. There are colour constancy illusions, but I will not believe that this is one of them until I see two people actually taking opposite sides. However, that is socially impossible to achieve, because there will always be enough people to perversely take the opposite side for the lulz. The wording of the original Tumblr post, and the buzzfeed and Wired articles, deliberately encourage this.
What colours do you see in the dress in this picture? Not in any other picture, or the real dress, but this, the original picture that started it all. Feel free to experiment with different monitors and ambient lighting (but not editing the picture) and to give details in comments.
If I saw only the text of the supposed controversy I would instantly diagnose it as a hoax for clickbait. There are colour constancy illusions, but I will not believe that this is one of them until I see two people actually taking opposite sides.
My introduction to this illusion was walking into the office yesterday morning, and having a co-worker show me the picture on their monitor while asking what colour the dress was. The office split fairly evenly between white & gold versus blue & black, and nobody seemed to be kidding/trolling.
The weird specificity of a couple of people’s experiences also suggested they were being serious. One person who originally saw white & gold, after skimming blog posts discussing/explaining the illusion, eventually said they could kiiiinda see how it could be seen as blue & black (but that could’ve been a social conformity effect). Another person saw the dress as white & gold on their computer screen, but saw it as visibly blue & black when viewing their screen’s image through someone’s smartphone, which presented it with lower brightness.
So I deem this legit. I think this illusion messes with people so successfully by turning the relatively well-known brightness/saturation illusion (as in the checker shadow illusion) into a colour illusion, by exploiting the facts that (1) very light blue looks white, (2) brown looks like yellow/gold or black depending on intensity, and (3) perceived intensity depends on the surroundings as well as the region of focus.
I see sky blue and bronze-ish brown, which I’d interpret as navy blue and black decolored by aggressive washing. I still voted “blue and black” as I’d still call clothes that color black in real life unless I’m being pedantic, and there’s no way I’d call them gold.
Here and here are two fragments of the picture I linked, expanded but otherwise unaltered, and saved in an uncompressed format. Compression artefacts from the original are clearly visible, but please ignore these and attend only to the overall colours.
Again, this question asks only about your experience of the colours, not any guesses you might make about what you would see if you were there, nor what colours you can convince yourself you might be able to see.
If I didn’t know where these came from and I was doing an XKCD color survey-like thing, I’d call the colors in the first bronze and mauve and those in the latter black and indigo. (I’m not a native English speaker.) I’d call the difference between the two blacks “striking” but not the difference between the two blues, so I picked “Just show me the results”.
I think it has more to do with the fact that optical illusions are generally human universals, i.e. all humans see the same thing. (Certain illusions may be susceptible to cultural influence, but I don’t think that really applies here—Buzzfeed commenters are all generally from similar demographics.) Given this, it’s really weird that some people are seeing one thing and other people are seeing something else.
Case in point: I see it as white-and-gold, and I’ve looked over it several times already, with no change. I am actually having difficulty imagining how anyone could perceive it as blue-and-black, despite being fairly certain that the people their claiming to see blue-and-black are not lying. What’s strange isn’t the illusion itself, but the extremely polarizing effect it has on people.
Yes, I agree that projecting “true color” onto the real world is a mistake. I’m not sure those commenters are actually doing that, though. I think your interpretation of the Buzzfeed argument is something like this:
“It’s white-and-gold!”
“No, it’s blue-and-black! How can you think it’s white-and-gold?!”
Whereas my interpretation, I feel, is slightly more charitable:
“I see white-and-gold!”
“No way, I see blue-and-black! Why are you seeing something different from what I’m seeing?”
In other words, I feel that the dicussion isn’t quite as full of fallacious reasoning as you seem to be making it out to be (in that it could interpreted in a different way that makes it about something other than the mind projection fallacy). Of course, I could be wrong. What do you think?
I actually trust your interpretation over mine. I haven’t read through it too carefully and sense that my frustration has interfered with my interpretation a bit.
My interpretation is that a sizable majority are being very serious. People from my coding bootcamp have been discussing it on Slack for a while now… and it’s embarrassing.
I’m sure that deep down most people know that it’s some sort of optical illusion, but there’s a difference between “If I really really really examined my beliefs, this is what I’d find” and “this is what I believe after taking 5 seconds to think about it”.
I really get the sense that the overwhelming majority doesn’t get the idea that “true color” doesn’t exist.
the overwhelming majority doesn’t get the idea that “true color” doesn’t exist.
Count me among them.
I define “true color” as the frequency mix of light together with its brightness. It’s perfectly well measurable—for example, I happen to own a device which will tell me what color it’s looking at. People in photography and design care about “true color” very much—they carefully calibrate their devices (monitors, printers, etc.) to show proper colors.
“In this sense” you defined dress.trueColor as undefined, so I still don’t see what you are talking about.
By the way, normally color is defined at perception point so it already includes both the lighting and the reflective characteristics of the object. It’s common to observe that something is color X under, say, sunlight, and the same thing is color Y under, say, fluorescent lights. Girls understand that well :-)
The crucial point is, I think, the “observer” argument. Even if you compress all the other parameters—lighting, reflective characteristics, etc.--what you’re left with is still a two-place function, such that if you pass in a different observer, a different result is returned. You’re free to define “true color” as whatever you like, but that’s not going to change the fact that some people might look at what you define as “gold” and see black instead—like, for instance, in the optical illusion we’re discussing in this thread—and then they’re going to argue with you. And then you might argue back, saying, “True color is the frequency mix of light together with its brightness!”, and then they’ll say, “Well, that’s not what I’m seeing, so explain that using your ‘true color’,” and so on and so forth, when really the only source of the argument is a failure to recognize that color perception differs depending on the person.
In short, you and adamzerner aren’t actually disagreeing about anything that’s happening here. Rather, you and he are defining the phrase “true color” differently (you as “the frequency mix of light together with its brightness” and adamzerner as “the visual sensation that, when perceived, maps in the brain to a certain word trigger associated with a color concept”). That’s all this whole argument is: an argument about the definition of a word, and those arguments are the most useless of all.
(Hence why Eliezer wrote this post. Seriously, this post is in my opinion one of the most useful posts ever written on LW; I’ve linked people to it more times than I can count. Why are people still making these sorts of elementary errors?)
First, as an aside, my disagreement with adamzerner isn’t about the definition of “true color”, he thinks such a thing just doesn’t exist at all.
Second, color is not a two-argument function, not any more than length or weight or, say, acidity. The output of the two-argument function is called perception of color.
Consider wine. One of it’s characteristics is acidity. Different people may try the same wine and disagree about its tartness—some would say the tannins mask it, some would disagree, some would be abnormally sensitive to acidity, some would have the wine with a meal which would affect the taste, etc. etc. And yet, acidity is not a two-argument function, I can get out the pH meter and measure—objectively—the concentration of hydrogen ions in the liquid.
While consumers might debate the acidity of a particular wine, the professionals—winemakers—do not rely on perception when they quality-control their batches of wine. They use pH meters and ignore the observer variation.
It’s the same thing with color. People can and do argue about perception of color, but if you want to see what the underlying reality is, you pull out your photospectrometer (or a decent proxy like any digital camera) and measure.
Professionals—people in photography, design, fashion—cannot afford to depend on observer perception so they profile and calibrate their entire workflow. Color management is a big and important thing, and it’s a science—it does not depend on people squinting at screens and declaring something to be a particular color.
Think about a photographer shooting a catalog for a fashion brand. In this application color accuracy is critical because if he screws up the color, the return rates for the item will skyrocket with the customers saying “it’s the wrong color, it looks different in real life than in the catalog”. And if that photographer tries to say that true color doesn’t exist and he just sees it that way, well, his professional career is unlikely to be long.
See—this, right here? This is what I mean by “argument about a definition of a word”. I don’t care what you think “color” is; I care if we’re talking about the same thing. If you insist on defining “color” as something else, we are no longer discussing the same topic, and so our disagreement is void. You are talking about one concept (call that concept “roloc”) and adamzerner is talking about another concept (call that concept “pbybe”).
So, does “roloc” exist objectively? Yes, and adamzerner doesn’t disagree with that.
Does “pbybe” exist objectively? No, because it’s a two-place function like I was talking about, and you don’t disagree with that.
So what’s our disagreement here, exactly? Are we arguing about how to define the word color? From your comment, specifically the portion I quoted above, I get the sense that to you, that is what we are arguing about. “Color is not x; it’s y.” Well, I say screw that. I’m not here to argue about definitions of words. You call your thing “color” if you want, and I’ll call mine something different, like “Bob”.
Because your previous comment showed that you were still engaging in an argument about the definition of a word, despite your claims to the contrary, and I was under the impression that you would appreciate it if I pointed that out. Clearly I was mistaken, seeing as your reply contains 100% snark and 0% content. I regret to say that this will be my last reply to you on this thread, seeing as you are clearly not interested in polite or reasoned discussion. Insulting snark does not a good response make.
Optical illusions might be a good example of this. Ex. http://www.buzzfeed.com/catesish/help-am-i-going-insane-its-definitely-blue#.hfZoPkgjK
People are freaking out over what color the dress “really is”. They’re projecting the property of “true color” onto the real world, when in reality “true color” is only in their mind.
The way I think about it:
Oh, dear Lord, this made it even to LW… X-/
Human language isn’t great for talking about colors. The word “blue”, for example, encompasses very different hues.
Let’s think about this in terms of the HSV (hue, saturation, value) model.
In this model, whitish colors have low saturation, high value (value is basically brightness), and the hue doesn’t matter. Similarly, blackish colors have high saturation, low value, and the hue doesn’t matter again.
The dress in this image has the blue hue, but high value (=brightness). That makes it light blue or bluish white—take your pick. The other color has the orange hue, but low value. That makes it dark brown (gold) or brownish black, again, take your pick.
Overlay on top of this the differences in uncalibrated monitors, personal idiosyncrasies, and some optical illusion effects, and you have a ridiculously successfull buzzfeed post :-D
Yeah, that’s a pale blue, but I find it hard to alieve (and, had I not read Generalizing From One Example, I’d also find it hard to believe) that a sizeable fraction of the population find it pale enough to call it “white” non-trollingly. This also applies if I look at it on a dark background such as the left panel of XKCD with my screen’s luminosity set to the maximum. (The only condition in which I would call such a thing “white” is if I thought it only looked like that because of the ambient light, but under so blue a light the dark parts of the dress would look quite black with hardly any trace of yellow whatsoever.)
I see the same shade of pale blue while walking to work over white snow.
Are they, actually? Or are they freaking out over what other people say it is?
If I saw only the text of the supposed controversy I would instantly diagnose it as a hoax for clickbait. There are colour constancy illusions, but I will not believe that this is one of them until I see two people actually taking opposite sides. However, that is socially impossible to achieve, because there will always be enough people to perversely take the opposite side for the lulz. The wording of the original Tumblr post, and the buzzfeed and Wired articles, deliberately encourage this.
What colours do you see in the dress in this picture? Not in any other picture, or the real dress, but this, the original picture that started it all. Feel free to experiment with different monitors and ambient lighting (but not editing the picture) and to give details in comments.
[pollid:831]
My introduction to this illusion was walking into the office yesterday morning, and having a co-worker show me the picture on their monitor while asking what colour the dress was. The office split fairly evenly between white & gold versus blue & black, and nobody seemed to be kidding/trolling.
The weird specificity of a couple of people’s experiences also suggested they were being serious. One person who originally saw white & gold, after skimming blog posts discussing/explaining the illusion, eventually said they could kiiiinda see how it could be seen as blue & black (but that could’ve been a social conformity effect). Another person saw the dress as white & gold on their computer screen, but saw it as visibly blue & black when viewing their screen’s image through someone’s smartphone, which presented it with lower brightness.
So I deem this legit. I think this illusion messes with people so successfully by turning the relatively well-known brightness/saturation illusion (as in the checker shadow illusion) into a colour illusion, by exploiting the facts that (1) very light blue looks white, (2) brown looks like yellow/gold or black depending on intensity, and (3) perceived intensity depends on the surroundings as well as the region of focus.
I see sky blue and bronze-ish brown, which I’d interpret as navy blue and black decolored by aggressive washing. I still voted “blue and black” as I’d still call clothes that color black in real life unless I’m being pedantic, and there’s no way I’d call them gold.
Here and here are two fragments of the picture I linked, expanded but otherwise unaltered, and saved in an uncompressed format. Compression artefacts from the original are clearly visible, but please ignore these and attend only to the overall colours.
Again, this question asks only about your experience of the colours, not any guesses you might make about what you would see if you were there, nor what colours you can convince yourself you might be able to see.
[pollid:832]
I wouldn’t go as far as to call it a striking difference, but I perceive the first image’s colours as lighter and more washed out than the second’s.
If I didn’t know where these came from and I was doing an XKCD color survey-like thing, I’d call the colors in the first bronze and mauve and those in the latter black and indigo. (I’m not a native English speaker.) I’d call the difference between the two blacks “striking” but not the difference between the two blues, so I picked “Just show me the results”.
I think it has more to do with the fact that optical illusions are generally human universals, i.e. all humans see the same thing. (Certain illusions may be susceptible to cultural influence, but I don’t think that really applies here—Buzzfeed commenters are all generally from similar demographics.) Given this, it’s really weird that some people are seeing one thing and other people are seeing something else.
Case in point: I see it as white-and-gold, and I’ve looked over it several times already, with no change. I am actually having difficulty imagining how anyone could perceive it as blue-and-black, despite being fairly certain that the people their claiming to see blue-and-black are not lying. What’s strange isn’t the illusion itself, but the extremely polarizing effect it has on people.
I definitely agree—it’s a particularly curious illusion. Some researcher seems to think so as well.
But the point remains that it can be explained by understanding the illusion, and that projecting “true color” onto the real world is a mistake.
Yes, I agree that projecting “true color” onto the real world is a mistake. I’m not sure those commenters are actually doing that, though. I think your interpretation of the Buzzfeed argument is something like this:
Whereas my interpretation, I feel, is slightly more charitable:
In other words, I feel that the dicussion isn’t quite as full of fallacious reasoning as you seem to be making it out to be (in that it could interpreted in a different way that makes it about something other than the mind projection fallacy). Of course, I could be wrong. What do you think?
I actually trust your interpretation over mine. I haven’t read through it too carefully and sense that my frustration has interfered with my interpretation a bit.
In fairness, they’re mostly kidding and exaggerating.
My interpretation is that a sizable majority are being very serious. People from my coding bootcamp have been discussing it on Slack for a while now… and it’s embarrassing.
I’m sure that deep down most people know that it’s some sort of optical illusion, but there’s a difference between “If I really really really examined my beliefs, this is what I’d find” and “this is what I believe after taking 5 seconds to think about it”.
I really get the sense that the overwhelming majority doesn’t get the idea that “true color” doesn’t exist.
Count me among them.
I define “true color” as the frequency mix of light together with its brightness. It’s perfectly well measurable—for example, I happen to own a device which will tell me what color it’s looking at. People in photography and design care about “true color” very much—they carefully calibrate their devices (monitors, printers, etc.) to show proper colors.
Go look at e.g. CIE color spaces.
I mean “true color” in this sense:
“In this sense” you defined dress.trueColor as undefined, so I still don’t see what you are talking about.
By the way, normally color is defined at perception point so it already includes both the lighting and the reflective characteristics of the object. It’s common to observe that something is color X under, say, sunlight, and the same thing is color Y under, say, fluorescent lights. Girls understand that well :-)
The crucial point is, I think, the “observer” argument. Even if you compress all the other parameters—lighting, reflective characteristics, etc.--what you’re left with is still a two-place function, such that if you pass in a different observer, a different result is returned. You’re free to define “true color” as whatever you like, but that’s not going to change the fact that some people might look at what you define as “gold” and see black instead—like, for instance, in the optical illusion we’re discussing in this thread—and then they’re going to argue with you. And then you might argue back, saying, “True color is the frequency mix of light together with its brightness!”, and then they’ll say, “Well, that’s not what I’m seeing, so explain that using your ‘true color’,” and so on and so forth, when really the only source of the argument is a failure to recognize that color perception differs depending on the person.
In short, you and adamzerner aren’t actually disagreeing about anything that’s happening here. Rather, you and he are defining the phrase “true color” differently (you as “the frequency mix of light together with its brightness” and adamzerner as “the visual sensation that, when perceived, maps in the brain to a certain word trigger associated with a color concept”). That’s all this whole argument is: an argument about the definition of a word, and those arguments are the most useless of all.
(Hence why Eliezer wrote this post. Seriously, this post is in my opinion one of the most useful posts ever written on LW; I’ve linked people to it more times than I can count. Why are people still making these sorts of elementary errors?)
I think you’re wrong.
First, as an aside, my disagreement with adamzerner isn’t about the definition of “true color”, he thinks such a thing just doesn’t exist at all.
Second, color is not a two-argument function, not any more than length or weight or, say, acidity. The output of the two-argument function is called perception of color.
Consider wine. One of it’s characteristics is acidity. Different people may try the same wine and disagree about its tartness—some would say the tannins mask it, some would disagree, some would be abnormally sensitive to acidity, some would have the wine with a meal which would affect the taste, etc. etc. And yet, acidity is not a two-argument function, I can get out the pH meter and measure—objectively—the concentration of hydrogen ions in the liquid.
While consumers might debate the acidity of a particular wine, the professionals—winemakers—do not rely on perception when they quality-control their batches of wine. They use pH meters and ignore the observer variation.
It’s the same thing with color. People can and do argue about perception of color, but if you want to see what the underlying reality is, you pull out your photospectrometer (or a decent proxy like any digital camera) and measure.
Professionals—people in photography, design, fashion—cannot afford to depend on observer perception so they profile and calibrate their entire workflow. Color management is a big and important thing, and it’s a science—it does not depend on people squinting at screens and declaring something to be a particular color.
Think about a photographer shooting a catalog for a fashion brand. In this application color accuracy is critical because if he screws up the color, the return rates for the item will skyrocket with the customers saying “it’s the wrong color, it looks different in real life than in the catalog”. And if that photographer tries to say that true color doesn’t exist and he just sees it that way, well, his professional career is unlikely to be long.
See—this, right here? This is what I mean by “argument about a definition of a word”. I don’t care what you think “color” is; I care if we’re talking about the same thing. If you insist on defining “color” as something else, we are no longer discussing the same topic, and so our disagreement is void. You are talking about one concept (call that concept “roloc”) and adamzerner is talking about another concept (call that concept “pbybe”).
So, does “roloc” exist objectively? Yes, and adamzerner doesn’t disagree with that.
Does “pbybe” exist objectively? No, because it’s a two-place function like I was talking about, and you don’t disagree with that.
So what’s our disagreement here, exactly? Are we arguing about how to define the word color? From your comment, specifically the portion I quoted above, I get the sense that to you, that is what we are arguing about. “Color is not x; it’s y.” Well, I say screw that. I’m not here to argue about definitions of words. You call your thing “color” if you want, and I’ll call mine something different, like “Bob”.
“Color” exists; “Bob” doesn’t. There, disagreement settled. Okay? Okay.
It’s cool how you talk to yourself, but why is this comment tagged as an answer to me?
Because your previous comment showed that you were still engaging in an argument about the definition of a word, despite your claims to the contrary, and I was under the impression that you would appreciate it if I pointed that out. Clearly I was mistaken, seeing as your reply contains 100% snark and 0% content. I regret to say that this will be my last reply to you on this thread, seeing as you are clearly not interested in polite or reasoned discussion. Insulting snark does not a good response make.
..
..
“Polite or reasoned discussion”, right… X-D