Cognitive Daily

The World Color Survey is a massive project which attempts to understand how colors are categorized in different languages. The researchers studied 110 different languages, none of which had a written component, which ensured that only spoken word categories would be used to describe the colors.

Do the speakers all understand colors the same way? Is “red” red whether you’re speaking Chumburu or Saramaccan? Rolf Kuehni undertook an analysis of the data to try to find out.

To discuss colors and language, it’s important to differentiate between the word we’re using to describe a color, and the color itself. Color researchers use the term “gloss” to distinguish between the two. So “red gloss” is the word being used to describe the color red (which might be “rouge”, “rosso”, or “red”), while “red” refers to the color itself (since we’re speaking English right now). In the World Color Survey, four hues were identified as unique — red, green, yellow, and blue.

This means that each of these colors tend to have unique words to describe them in each language. For example, in English, though we distinguish between light green and dark green, we understand that both colors are also green. In this sense, only the green gloss is unique. We never use the blue gloss to describe the color green.

Forty-one of the 110 languages in the survey have words corresponding to each unique hue. In each language, the range of colors identified as corresponding to the gloss for each color spans nearly the identical range of hues. Here’s a graph summarizing those results for 38 of the languages (3 were eliminated from analysis because too few participants spoke those languages):

The numbers along the bottom of the chart correspond to physical color chips that researchers used to display colors. Consider the yellow gloss. Once it has been determined that a particular language has a word for “yellow,” researchers asked observers speaking to pick the color chip that best represents that color. As you can see, for each language, the range of colors chosen was nearly identical.

But what about the other 69 languages? Their results actually followed a similar pattern in most cases. For example, 40 of the languages had a term corresponding to “grue” — a mixture of green and blue. These languages still broke the colors into similar categories, but lumped two of the categories together. Even among languages that had all four color categories, “blue” and “green” were the most frequently confused colors. Other languages had words for “grelow” and “relow,” but overall, there was close to 90 percent agreement on the boundaries of color categories.

Kuehni argues that this is evidence for common color categories in all humans. He’s careful to point out that while this doesn’t rule out the possibility that colors are culturally determined, given the wide range of cultures studied here, it’s strong support for the idea that color categories are determined neurophysiologically — that the way we categorize colors is embedded in the structure of the brain.

Kuehni, R.G. (2005). Focal color variability and unique hue stimulus variability. Journal of Cognition and Culture, 5(3-4), 409-426.

Comments

#1 thinker
September 11, 2006

How would things like red-deficiency (sizeable percentage of males) and similar differences play into color perception, do you think?
#2 Sandeep Gautam
September 12, 2006

How does this relate to earlier work on color language terms that had resulted in a heirarchical ordering of languages based on the color terms used in those languages. Was this study restricted to stage V languages (the 41 languages) and stage IV languages (the rest 69 languages)? For some more info on the neurophysiological grounding of color categories, my blog postings on the same may be informative.
#3 John Wilkins
September 12, 2006

Or, it’s evidence that we all share the same rods and cones in our retinas, and that they activate at the same general frequencies. How much of this is due to differences in biology and differences in cultures is open to question, I would have thought.
#4 Dave Munger
September 12, 2006

John: If the color categories were due to the cones and rods in our eyes, then wouldn’t the four unique hues correspond to those colors: red, green, and blue? Actually, their values are closer to yellowish green, green, and bluish violet, which certainly bears little resemblance to the unique hues.

Thinker: In this study, I believe “color-blind” individuals were excluded from analysis. Since color blindness is relatively rare (about 10 percent of the male population), I wonder if it would have much effect at all on cultural color norms. I’d be interested to hear of any research suggesting it does. One thing I can point to from this study: the most common type of color-blindness is red-green (observers have difficulty distinguishing between red and green). But red and green are not commonly confused in this study — more common is a blurring of green and blue, or green and yellow: adjacent rather than opponent colors on the color wheel.
#5 stan
September 12, 2006

Why is he careful to point out that this does not rule out the possibility that colors are culturally determined? It seems to me that the experiment presupposes that we think of the colors that the various “color glosses” latch onto as culture-independent (as well as “mind-independent”). How else are we able to say things, during the experiment, like “they refer to greenish-blue with their ‘grue gloss’”?

I don’t see how the issue of whether or not color “categories” (“the WAY we categorize colors”) are determined neurophysiologically is relevant to the issue of whether colors (themselves)are so determined?
#6 TLTB
September 13, 2006

If indeed we have some kind of “color faculty” of the mind that causes us to acquire colors in certain categories, one wonders if there would be an evolutionary advantage to developing this faculty; furthermore, there should be an advantage to gaining the specific categories that the study found we (universally) have. It may be significant that the categories are kinds of green and blue – colors that occur most frequently in the natural environment.
#7 Dave Munger
September 13, 2006

Good question, TLTB — one I can’t answer. I’m not sure the point about green and blue holds up, though. The most frequently categorized color is red, followed by yellow. Green and blue are the most frequently confused. But you may be right still, in that blue and green are the most common in nature: it would be important to identify *uncommon* colors, such as poisonous red berries or a yellow lion.
#8 Jerry Schwarz
September 14, 2006

In 67 I was a sophomore in college with several years of
programming experience.I got a job writing a program for a psychology
professor. He was trying to find color descriptions that could be used
to reliably distinction groups of items. For example you give
seven differently colored wires to somebody and ask them to pick the
red one. He had collected some data on what selection people made in
this kind of situation. My program was supposed to search his data
and find a maximum collection of color descriptions that could be
reliable distinguished. I naively thought this would be easy to do.
I did write the program but, as you might guess, it hit combinatoric
issues and I never managed to get it to find more than four or five
colors. In retrospect I suspect it was an NP hard problem (although
this was several years before that notion was formulated).

I realize this comment is only peripherally related to the post,
but I wonder if anyone knows whether this line of research has
been pursued, and if so what the state of our knowledge is.
#9 David Harmon
September 27, 2006

thinker: Red-green color-blindess isn’t quite [i]that[/i] common. IIRC, it’s 7% of males in America, but that probably includes some immigrant groups where it’s favored. In any case, color-blind people pragmatically have no influence on general color naming, because (1) they’re a small minority, and (2) with respect to color, it’s a disability which even they eventually have to recognize. An example I’ve used before: my color-blind father would ask my stepmother which suit jacket was brown vs. blue, because if he went out mismatched, “looks the same to me” wouldn’t cut the mustard at court. Triple that for traffic lights!

Correspondingly, the (rare) women with four-color vision would have little opportunity to coin new “glosses” outside the standard color-space, because everyone else (except maybe relatives sharing the condition) would insist on considering the new word as a neologism for a familiar color.

Jerry Schwarz: Well, the reviewed article itself is similar, though it approaches the issue from the flip side. In general, if I, a programmer born about the time you entered college, were to try something like that, I wouldn’t even bother with exhaustive combinatorics, I’d feed the data to one of the modern iterative analysis schemes, such as a (simulated) neural network or even a genetic algorithm.