July 21, 2013 at 20:23xkcd presents: color survey results
- Transfer
Who, looking at the rainbow, is able to indicate the exact border, where the blue tones end and the orange begin? We clearly see the difference in colors, but where is one replacing the other? This is also the case with diseases of the mind.
- Herman Melville, “Billy Bad, Fore-Mars Sailor”
Orange? Red? I don’t know what to believe!
- Anonymous, “A Poll of Flowers”
IF THE NEXT QUESTION WILL BE AGAIN ABOUT COLORS, I EXPRESS YOUR HEART WITH A SPOON AND EAT ITS HIM
- Anonymous, “A Poll of Flowers”
Many thanks to all who participated in the poll! It was processed 222 thousand profiles containing more than five million color names. Answers are no longer accepted (as early as three years) , but if you are curious, here are the questions and instructions:
For starters, some interesting facts:
All in all, I got a lot of pleasure analyzing your answers. I liked the results too; even despite the fact that this “polling” approach has its limits (which will be discussed at the end of the post), the sheer amount of data collected is fascinating.
By a strange coincidence, on the same night when I opened the survey for public access, the following picture appeared on the Doghouse Diaries website :
I thought that now I can check how true this is. After the survey came to an end, I made my version using the data collected:
In fact, women just use prefixes (such as “bright”) a little more often, but otherwise they agree with men (and some other differences can be written off sampling imperfection).
Then I decided to define the “most masculine” and “most feminine” colors. That is, I looked for colors that men often called, and women rarely (and vice versa).
Here is a list of the “most feminine” colors:
A magnificent set, what else can I say. The incense bomb planted in Bed Bath & Beyond immediately appears. Okay, now let's look at the second list.
"The most male" colors:
I ... by the way, this is not my typo in the fifth paragraph - the only color on this list is the word "beige", typed with an error. This can not be attributed to a dozen illiterate users - I took into account the number of people who named the color, and not how many times the color was named. And this is after the screening of spammers.
In a word, I am ashamed of the entire male gender. But it's time to move on:
I managed to pretty accurately determine the RGB values of about a thousand colors. Here are the first 48 of them:
A complete table of 954 colors is available here , or as a text file . I'm not sure about whether to use this data as the new X11 rgb.txt . Most likely, the transition will be very painful.
The choice of the RGB value for the color is based on the coordinates of the point in the color space at which the distribution density of the answers for this color was maximum. I tried to just take the geometric mean (which, in principle, was wrong), sort out the coordinates of points suitable for the role of the "center" (too slow), adjust the nuclear density functions (too hard), but in the end I just launched it several timesa randomized search with an ascent to the top and took an average value. If anyone is interested in the details, they are listed at the end of the color table.
Many people have spelling problems:
Note that the correct option is missing. The fact is that I also couldn’t spell this word correctly and relied on Google when I made up the tablet:
But now I know a simple rule: you can just mentally break the word into pieces - “fuch-sia” (almost like “fuck-sia” )
In addition to errors, there were frank attempts to spam the survey, which I fought in several ways. Firstly, I did not take into account those people who entered too many “rare” names (that is, those who did not appear in the answers of other people). Secondly, users were discarded, who equally named several colors that differed greatly in shade (most often, they entered the same color again and again). Some have written scripts to facilitate their work; the winner was (judging by the filters) a man who named 2400 flowers with one short racist insult.
Part of the survey (1.5 million results) was limited to three faces of the RGB cube. Here is a map of this region with drawn borders between colors:
You can also download a larger map , a full-scale map or data in a text format .
As I said, data obtained in this way cannot be relied on completely. For example, previous colors influence how a person perceives the next color. Moreover, the color rendering of the screens is different; RGB is not an absolute color space. Fortunately, I was only interested in how the colors looked on the average monitor, so I only analyzed the responses of users who indicated that they use a regular monitor (LCD - more than 90%, CRT - 6%) and do not suffer from color blindness.
In general, color is a very exciting subject to study, especially considering the fact that each of us has a stock of disparate (and often contradictory) knowledge about color perception, light waves, primary colors, color models, and rainbows. If you are interested, I advise you to read Feynman's lectures on physics (vol. 1, chap. 35), Charles Poynton's Color FAQ (in English), or wander around the links on Wikipedia .
Finally, here are some unusual colors named during the survey.
... For the most part, poorly translatable. I apologize in advance to the unknown boy, next to whom I had to put the ugly word “sweatshirt”, as well as to all those readers who know English.
If you have your own ideas on how to analyze this data, they are available here in the form of SQLite-base. I will be very happy to know what happened to you; write to me at xkcd@xkcd.com (in English, please) .
- Herman Melville, “Billy Bad, Fore-Mars Sailor”
Orange? Red? I don’t know what to believe!
- Anonymous, “A Poll of Flowers”
IF THE NEXT QUESTION WILL BE AGAIN ABOUT COLORS, I EXPRESS YOUR HEART WITH A SPOON AND EAT ITS HIM
- Anonymous, “A Poll of Flowers”
Many thanks to all who participated in the poll! It was processed 222 thousand profiles containing more than five million color names. Answers are no longer accepted (as early as three years) , but if you are curious, here are the questions and instructions:
- What is your monitor?
- I do not know
- LCD (flat panel or laptop screen)
- CRT (large box with a glass screen)
- other (projector, e-ink, ...)
- What is the color temperature of your monitor? If you don’t know, select the “I don’t know” option or try to find out in the monitor settings.
- What is the contrast ratio (gamma) of your monitor?
- Do you distinguish colors?
- Do you have a Y chromosome? If you are not sure, select “yes” if you were male at birth, and “no” if female.
- What is your native language? (Whatever it is, please enter the color names in English.)
On each survey page, enter the name of the color that you see, and press Enter. You can answer both clearly and vaguely; “Green” and “light olive” are equally good. You can also give the same answer several times.
The survey does not have a pronounced end. You can answer many questions at a time, take a break and continue later. Share the link with your friends so they can participate too. The main thing - do not confer.
It's okay if you made a typo; just keep answering. The back button does not return you to the previous color, but only leads to the fact that several answers remain unaccounted for.
For starters, some interesting facts:
- If you force people to call colors long enough, they become violent.
- Whatever one says, “vomit” (puke) and “vomit” (vomit) are also colors.
- Color blind people are more likely than non-color blind people to enter "fuck this" and stop responding.
- As many expected, the “blue” (indigo) was included in the rainbow only so that there were seven flowers, and also because every hunter wants to know where the pheasant is sitting (ROY G. BIV). In fact, every hunter wants to know where the pub is (ROY GBP), the giant hare where to shoot (ROYGTBP) or even the king where to meet (ROYGCBP).
- About twenty people have inserted the 'drop table' SQL code into their answers. Great try guys keep up the good work.
- No one could spell the word "fuchsia" (fuchsia) correctly.
All in all, I got a lot of pleasure analyzing your answers. I liked the results too; even despite the fact that this “polling” approach has its limits (which will be discussed at the end of the post), the sheer amount of data collected is fascinating.
Floor
By a strange coincidence, on the same night when I opened the survey for public access, the following picture appeared on the Doghouse Diaries website :
I thought that now I can check how true this is. After the survey came to an end, I made my version using the data collected:
In fact, women just use prefixes (such as “bright”) a little more often, but otherwise they agree with men (and some other differences can be written off sampling imperfection).
Then I decided to define the “most masculine” and “most feminine” colors. That is, I looked for colors that men often called, and women rarely (and vice versa).
Here is a list of the “most feminine” colors:
- faded greenish blue (Dusty Teal)
- blush (blush pink)
- faded lavender (Dusty Lavender)
- butter (Butter Yellow)
- dull rose (Dusky Rose)
A magnificent set, what else can I say. The incense bomb planted in Bed Bath & Beyond immediately appears. Okay, now let's look at the second list.
"The most male" colors:
- Penis
- Gay
- Wtf
- don't know (Dunno)
- beige (Baige)
I ... by the way, this is not my typo in the fifth paragraph - the only color on this list is the word "beige", typed with an error. This can not be attributed to a dozen illiterate users - I took into account the number of people who named the color, and not how many times the color was named. And this is after the screening of spammers.
In a word, I am ashamed of the entire male gender. But it's time to move on:
RGB
I managed to pretty accurately determine the RGB values of about a thousand colors. Here are the first 48 of them:
A complete table of 954 colors is available here , or as a text file . I'm not sure about whether to use this data as the new X11 rgb.txt . Most likely, the transition will be very painful.
The choice of the RGB value for the color is based on the coordinates of the point in the color space at which the distribution density of the answers for this color was maximum. I tried to just take the geometric mean (which, in principle, was wrong), sort out the coordinates of points suitable for the role of the "center" (too slow), adjust the nuclear density functions (too hard), but in the end I just launched it several timesa randomized search with an ascent to the top and took an average value. If anyone is interested in the details, they are listed at the end of the color table.
Spelling and spam
Many people have spelling problems:
Note that the correct option is missing. The fact is that I also couldn’t spell this word correctly and relied on Google when I made up the tablet:
But now I know a simple rule: you can just mentally break the word into pieces - “fuch-sia” (almost like “fuck-sia” )
In addition to errors, there were frank attempts to spam the survey, which I fought in several ways. Firstly, I did not take into account those people who entered too many “rare” names (that is, those who did not appear in the answers of other people). Secondly, users were discarded, who equally named several colors that differed greatly in shade (most often, they entered the same color again and again). Some have written scripts to facilitate their work; the winner was (judging by the filters) a man who named 2400 flowers with one short racist insult.
Map
Part of the survey (1.5 million results) was limited to three faces of the RGB cube. Here is a map of this region with drawn borders between colors:
You can also download a larger map , a full-scale map or data in a text format .
Main disadvantages
As I said, data obtained in this way cannot be relied on completely. For example, previous colors influence how a person perceives the next color. Moreover, the color rendering of the screens is different; RGB is not an absolute color space. Fortunately, I was only interested in how the colors looked on the average monitor, so I only analyzed the responses of users who indicated that they use a regular monitor (LCD - more than 90%, CRT - 6%) and do not suffer from color blindness.
In general, color is a very exciting subject to study, especially considering the fact that each of us has a stock of disparate (and often contradictory) knowledge about color perception, light waves, primary colors, color models, and rainbows. If you are interested, I advise you to read Feynman's lectures on physics (vol. 1, chap. 35), Charles Poynton's Color FAQ (in English), or wander around the links on Wikipedia .
All sorts of things
Finally, here are some unusual colors named during the survey.
... For the most part, poorly translatable. I apologize in advance to the unknown boy, next to whom I had to put the ugly word “sweatshirt”, as well as to all those readers who know English.
If you have your own ideas on how to analyze this data, they are available here in the form of SQLite-base. I will be very happy to know what happened to you; write to me at xkcd@xkcd.com (in English, please) .
ColorDB- data from the main part of the survey, SatOnly- from the one in which only the colors from the faces of the RGB cube were shown, on a black and white background.