Where does “Blue” end and “Red” begin?
March 17th, 2008 by Brendan O'ConnorWhat would you call these colors?
We showed thousands of random colors like this to people on Mechanical Turk and asked what they would call them. Here’s what they said:
The above picture contains about 1,300 colors and the names for them that Turkers gave. Each is printed in its color and positioned on a color wheel. Just looking around, there sure seem to be different regions for different names. But there are also rich sets of modifiers (”light”, “dark”, “sea”), multiword names (”army green”), and fun obscure ones (”cerulean”). To help look at all this, we also made a color label explorer, so you can search for different terms and see different parts of the space. If the link doesn’t work for you, here are a few examples:
 |
 |
 |
 |
This study is basically the same design as the famous World Color Survey, where anthropologists showed color patches to speakers of many different languages and asked for names, to test the universality of language. Of course, we have mostly native English speakers. However, we can get much more data. (The above picture and links use only a small percentage of all the colors and names we collected.) There’s tons more that can be done. Want to make a better visualizer? Statistical analysis of colors to name terms? Let us know and we should be able to get this data set online.
UPDATE 3/18: I posted the data set.
March 17th, 2008 at 11:24 pm
cerulean obscure?? obviously no painters on staff as it’s a standard color in oils and acrylic paints.
March 18th, 2008 at 12:08 am
I wonder how much variation would disappear if all the respondents were using correctly (or at least identically) calibrated monitors. Neat experiment, though! My two-year-old son and I frequently disagree on whether one his toys is blue (him) or purple (me). Mom agrees with him, FWIW.
March 18th, 2008 at 12:27 am
Brendan,
I for one would love to see the dataset made available. One interesting question: on what colors were there most agreement? Least?
March 18th, 2008 at 12:29 am
Hi, I’d love to play with this dataset. Depending on what is in it, you could do interesting things like look at the relationship between the colors presented to Turkers and their names, whether certain colors took longer to label than others, which colors have a higher level of “agreement” than others, etc.
March 18th, 2008 at 12:31 am
Baby poop? Ewww. Very neat though.
March 18th, 2008 at 12:37 am
I just finished teaching a class in color exploration at the Virginia Museum of Fine Arts Studio School. I’m definitely going to share this with my students. I think it’s fascinating.
March 18th, 2008 at 1:26 am
No printers either: no “reflex blue” or “process blue.”
March 18th, 2008 at 1:43 am
When u mix red with blue? Whatever…
March 18th, 2008 at 1:44 am
I wonder how many of the respondents were colour blind? Stephan, it would be worth getting tested - I’m protanopic and I have the exact same problem.
March 18th, 2008 at 3:44 am
Do you have gender info for the namers? I’d like to see that.
March 18th, 2008 at 3:46 am
Neat!
linked on boingboing, btw
March 18th, 2008 at 4:17 am
I’m building a new cross-over colour theory course for studio artists and science students. I love this, and I’d like to cast another vote for getting access to the dataset. (Ditto on cerulean and process blue–your Turks were general audience, not colour professionals.)
March 18th, 2008 at 8:06 am
I fell out of a window
A second story window
And caught my eyelash on the sill.
March 18th, 2008 at 8:55 am
Thanks for all the feedback folks.
“cerulean obscure” –>my fault. Actually Lukas and Chris are both painters, I feel chagrined on this :)
Monitor calibration –> I’m sure it affects things substantially. With all the variation from that, plus variance between individuals and the incredible subjectiveness of the task, I’m just happy to see stable regions on the color space!
The data set –> I’m working on releasing it, check back soon!
March 18th, 2008 at 8:58 am
@Linda Carston:
Yes, the mechanical turks were probably a general audience. Isn’t that the point?
March 18th, 2008 at 8:58 am
Correction to above: Carson, not Carston. My apologies.
March 18th, 2008 at 2:25 pm
Cerulean is also a standard color in the Crayola 64 crayon box; I’m actually surprised there aren’t more Crayola entries on the palette. (Lazer Lemon, anyone?) Turkers must have long since outgrown the coloring books!
March 18th, 2008 at 2:29 pm
I use two monitors - one laptop and other LCD flat and it’s a bugger to get the colors sync’d up.
Good article, found another one for you if you’re interested:
Top 10 Awesome Websites That Sell Cool Products You Probably Have Never Visited But Need To.
http://www.comember.net/blogs/firepixel/
Take care
March 18th, 2008 at 3:23 pm
I think this is a terrific study. i also would like to see the dataset and also to repeat the experiment among many different groups. A university campus would be great and to get a little bit of background on the responders could be a riveting socio-anthro sketch.
March 18th, 2008 at 3:24 pm
P.S. found this through boingboing.net FTW
March 18th, 2008 at 3:44 pm
By labeling with something that his long horizontally and short vertically, you’ve created a horizontal blur effect. Thus the green wedge seems broader than the blue wedge.
March 18th, 2008 at 3:56 pm
Wonderful fun. I’d think presenting people with several colors side-by-side would have created more drift than varying monitors, as we perceive colors relative to each other. Much better to show one color at a time. Of course I didn’t see the survey itself, so I can’t be sure how you presented it. BTW, it seems to me you’ve already given us the data set; a ‘view source’ on the color label page pretty much covers it. Are you okay with us remixing it?
March 18th, 2008 at 5:07 pm
“My two-year-old son and I frequently disagree on whether one his toys is blue (him) or purple (me). Mom agrees with him, FWIW.”
You probably have a type of color blindness called anomalous trichromacy, of which there’s a subset called deuteranomalous which shift some colors slightly toward the red. It’s sex-linked, it effects mostly males, and causes a slight alteration to the color receptors in the eye (cones). These alterations can cause certain shades of blue to appear purple.
It’s not particularly rare, and certainly not harmful. Most people don’t even notice until, like you, they argue over the color of an object.
March 18th, 2008 at 5:23 pm
Hulkcredible!!!
March 18th, 2008 at 5:33 pm
Stephan (#2) wonders how much variation would disappear if all the respondents were using calibrated monitors, and I wonder how much would disappear if misspellings (e.g., “flourescent” instead of “fluorescent”) were corrected and capitalization/spacing were made consistent. Cool exercise, regardless.
March 18th, 2008 at 5:38 pm
For a great book on color (human vision, color perception, ontology, phenomenlogy, …) see “Color for Philosophers: Unweaving the Rainbow” by C. L. Harden.
March 18th, 2008 at 5:57 pm
You can tell how much of an impact Crayola has had on our country. I wouldn’t be surprised to find “macaroni and cheese” or “tickle me pink” up there…
March 18th, 2008 at 6:14 pm
“Statistical analysis of colors to name terms?”
I’m definitely interested!! I majored in Stats. I’d love to poke around with the data in Excel! :D
March 18th, 2008 at 6:14 pm
Could you make the source data available? I think it’d be a lot of fun to turn it into an interactive widget, if I got a few free moments. So you could type in a few letters and see examples like you give in real-time.
March 18th, 2008 at 6:24 pm
papayoung, actually you can look at an example of the survey used by following the link on “asked” right after the color boxes. OK, it’s a little hidden :) so here it is: http://s3.amazonaws.com/lab20/ae18a3141e0cd78369c043fda22c76bdd354ff48.html
Andrea, I tried looking at a few misspellings. For example, both “fushia” and “fuschia” appear a lot… “fusc?hia” gets both in the explorer.
March 18th, 2008 at 6:32 pm
To all interested — I finally got the data set online!
http://blog.doloreslabs.com/?p=17
March 18th, 2008 at 6:38 pm
The front porch of the house I grew up in was painted blue. Sure, it was a smoky, grayish sort of blue, but I always gave people directions by saying to look for “the house with the blue porch”. When my mom told me the porch was gray, I was very confused.
March 18th, 2008 at 6:40 pm
You say that each name is “printed in its color.” That would have required as many different ‘inks’ as there were different names offered (modulo synonymy) and also a procedure to decide which color ink went with each name. Didn’t you mean rather that you started with a version of color space and selected a bunch of points from that space for naming, then recorded each name at the point it was used to name? Apologies if this comes across as nit-picking, but I was genuinely confused at first.
March 18th, 2008 at 6:43 pm
I’ve often wondered if there are regional differences. Obviously humans learn color names from their parents and caregivers, largely when they’re rather young. So if we take a certain shade of blue-green, for instance, is it true that most people growing up in France will call it “blue” while most people growing up in the USA will call it “green”? That’s a made-up example, but there OUGHT to be such differences, and it would be interesting to explore them.
March 18th, 2008 at 7:16 pm
http://blog.doloreslabs.com/?p=11
Stephan & Philipp: I’ve got that problem where I confuse:
blue and purple
red and green
green and brown
pink and purple and violet (I’m pretty sure this is just disagreement over which word represents which colors)
As for the dataset projects I could REALLY REALLY USE a tool where I give the hex/rgb/hsv/cmyk value of a color and it gives me the words most people would use to describe that color or the color closest to it. There already is a tool similar to that named WhatColor but it’s Windows only and its color set is quite limited.
I’ve tried expanding the color set but the only data I’d found to work with before were fancy-pants tootie-fruity names of colors made by textile, interior decorating, and paint companies.
March 18th, 2008 at 7:47 pm
Lukas and Chris:
Nice work.
Are you aware of the NIST list of 500+ color names?
http://fire.nist.gov/fds/color_table.html
I just added it to my Social Bookmarks
http://del.icio.us/NetStrider?settagview=cloud
What you need is a random sample of adults who
work with colors to vote on each many color names.
Regards, George
March 18th, 2008 at 7:50 pm
This is a great concept for a statistic-mash-up web application. What would people call this color in China or Europe. What would women call this color if they are between 20 and 40 and so on.
Besides that it seems that you forgot the shades of white that would occur when additively mixing the colors.
Great idear!
March 18th, 2008 at 8:42 pm
Prof. Kay — by “printed in its color”, I just meant, when drawn on screen for that color wheel visualization. You’re right, all I did was start with a bunch of points in the color space and then showed them to people.
To be even more specific, the rendering was through whatever RGB system that people’s computers used (probably sRGB, I think but am not sure); and as others pointed out, differences in people’s monitors probably made a non-trivial difference in what the actually saw.
March 18th, 2008 at 9:09 pm
I think “Cubicle Gray” has a nice ring to it, as does “Pink Slip Rouge,” which is what color you’ll be getting if you read too much of MyLifeInaCube.com
March 18th, 2008 at 9:44 pm
I would like to see this experiment with technical parts of a computer or vehicle. with non-professionals. like in this study. “uhm a thing”, “oughta be a screw”, “somekindof screw”, “a piece of metal”, “definitely metal”.
March 18th, 2008 at 10:48 pm
i love the color blue,it’like summer sky,dawn.it’s also the color most used in hospital:(
March 18th, 2008 at 10:53 pm
i wonder if this was a random survey.Most of the people i know like the color blue.I’ts really irrelevant anywaythis is not one of my fav sites.i just thought i would put in my two cents,anyway.hehe
March 18th, 2008 at 11:48 pm
I like the one color blind green on there.
March 19th, 2008 at 4:24 am
What about anti-freeze green?
March 19th, 2008 at 8:08 am
Being color blind, I would love to see a survey of the colorblind and check out that representation. Maybe a question on the next survey.
March 19th, 2008 at 4:56 pm
Hola, hablar de colores es un tema muy complicado tantas gamas, tonos, degradados que creo que seria imposible identificar uno de otro, pero que hermosos son en la naturaleza creo que todavia los podemos apreciar.
March 19th, 2008 at 8:00 pm
Nice work, Brendan. If I recall correctly (a big if), the procedure that was used in the basic color term research by Kay and others, your procedure is the “inverse” of theirs: they started with terms (basic color terms) and asked which color chips they covered whereas you’re starting with a color chip and asking which terms cover them.
March 20th, 2008 at 4:49 am
Wy not a numbering system for colors instead of “fashion” names?
March 20th, 2008 at 3:09 pm
Apart from colour blindness in males, women also have a greater range of colour perception than men - up to 6 colour receptors instead of the usual 3 - and greater variation.
March 20th, 2008 at 5:26 pm
Brendan, it seems like no one else shared my confusion, so let’s forget it. BTW, cool stuff.
Stuart Robinson, you’re right that in the earliest work (Basic Color Terms. Berlin & Kay. 1969) we gave people names and asked them to indicate the corresponding color chips (in a fixed array), but in the later World Color Survey we presented color chips individually for naming. Check out http://www.icsi.berkeley.edu/wcs/ if you’re interested. For one thing, you’ll see the palette of chips used, and we still used that on the “show me the best example of …” task.
March 22nd, 2008 at 4:34 am
Just failed some color-tests so my flight license (in the UK/EU) will be restricted to flight during day time. The test were named Ishihara plates, Beynes latern test and Holmes Wright latern test. Never new about these test and neither about having a color vision deficiency/deviation. Some other approved test are available at flight authorities in switzerland, Germany and France. If one passes one of those (or USA or Australian ones) one can fly unrestricted within EU-airspace!
— I wander what is most confusing: some slight color deviations or EU-regulations…
March 23rd, 2008 at 6:47 am
Babypoop color? heheheh
March 24th, 2008 at 3:22 pm
Couldn’t find baby-shit green!
March 25th, 2008 at 4:05 am
If you liken the Color Label Explorer to a map of China and Mongolia, you get Tibet if you type in “blue,” Mongolia if you type in “green,” and Xinjiang (homeland of the Uyghurs) if you type in “aqua.” “Red” will get you that southeastern region around Guangdong and Hong Kong.
I know, it’s kind of a strange way to look at it, but I’ve had kind of a strange day.
March 25th, 2008 at 9:11 am
George said “Are you aware of the NIST list of 500+ color names?
http://fire.nist.gov/fds/color_table.html”
That’s not quite what you think - the naming style and a least a large number of the names appear to be based on the X11 color name list, which was itself not very scientifically constructed. It’s also based on RGB device-coordinates, which isn’t very stable.
I worked up my own table in 1994 or so, yielding http://www.talisman.org/~erlkonig/software/pub/xcms-rgb.xcmsdb with about 10000 colors based on variants of 24 basic hues with 2 more between each of those (so orange, orange-goldenrod, goldenrod-orange, goldenrod, goldenrod-yellow, etc), 9 steps towards white (pastels), 12 steps towards black, and 4 steps towards gray. And the whole X11 database, some symbolically-named colors, and a few extra. Using RGBI (emission- instead of device-coordinates, to some extent) allow for using CMS for color correction, so that 0-255 had subjectively even steps instead of tending to be bunched near the brighter end.
Anyway, I’m thinking reviewing my 24 hue names in light of the research set should be interesting, but perhaps the results of my project from back then would be useful to someone else :-)
March 26th, 2008 at 10:55 pm
I have a friend who is blind. If some of the extraneous colors were deleted and this could be printed out somehow for a blind person, this would be very, very useful.
May 25th, 2008 at 4:29 am
Here’s a 3D visualizer for the data and some screenshots:
http://www.box.net/shared/qbunsqy0og
The script requires Python and the Panda3D engine. It uses the modified data set (data2.csv) posted by Nick. Use the left mouse button to pan and the right mouse button to zoom. It’s pretty easy to change the scale and orientation of the axes in the code. Improvements are welcome.
July 10th, 2008 at 9:12 pm
cool site , thanks