link to original article

The Tsimane’ people of the Amazon (pronounced chee-MAH-nay, roughly) hunt, farm, and forage. They don’t have a lot of technology. And if you talk to them about the colors they see in the world, they say some pretty interesting things.

For objects that are what an English speaker would probably describe as white, black, or red, the Tsimane’ have specific words too. But show them just about any other color, and their language multiplies. “There’s a huge amount of variability,” says Ted Gibson, a language researcher at MIT who works with the Tsimane’. “Different people would use different terms for what we would call blue and yellow and orange. They really divide up the space differently, depending on the person.”

So, like, Tsimane’ speakers who also speak a lot of Spanish say “yushnus” for blue (-ish) and “shandyes” for green (-ish). But others call both the sky and grass “yushnyes,” and others call all green and blue things “shandyes.” Or, to take another example, different Tsimane’ speakers call yellowish colors “cuchicuciyeisi,” “ifuyeisi,” or “chamus.”

In the world of color research, that’s unusual to the point of uniqueness. Across languages and cultures, people tend to break up “colorspace,” the universe of all the colors humans see, in roughly the same way—different words, sure, but for mostly the same colors. So when Gibson told Bevil Conway, a vision researcher at the National Institutes of Health’s Intramural Research Program, about his findings with the Tsimane’, Conway wanted to know more.

Conway studies color and vision; Gibson studies language and information theory, and had been looking at the way Tsimane’ speakers use numbers. Together, they cooked up a new research project to figure out how Tsimane’ speakers see what they see—or, more accurately, how they talk about what they see. Because if you accept that all humans with normal vision are capable of seeing the same vast range of colors (which they are) but vary in the way they talk about those colors (which they do), that says something about how they think. Color turns out to be a very good vector for figuring out how the human brain actually works.

The paper that Conway, Gibson, and their colleagues eventually wrote came out in last week’s Proceedings of the National Academies of Science. And the results are … illuminating. Gibson and Conway have a new theory about how the human brain makes color, and it is less about how people see and more about what they do.

In 1969 two UC Berkeley researchers named Brent Berlin and Paul Kay pulled an array of 320 color chips from the thousands that represented the work of Albert Munsell, a Boston artist and teacher who first proposed organizing colors based on hue (the wavelength of the photon), value (brightness), and chroma (saturation, like from pastel to vivid). Berlin and Kay showed their color array to 20 bilingual Bay Area residents, asking them, essentially, to name the colors in their native languages.

For Berlin and Kay, how people used words to describe points in Munsell’s three-dimensional “color space” was a chisel to use against what’s called the Sapir-Whorf hypothesis, an idea from linguistics that basically says language and culture influence what people think. If you don’t have a word for it, you can’t think about it—no word for “green” and you literally cannot see green, to be a reductionist Whorfian about it. So different cultures would necessarily have lots of different color words.

Berlin and Kay didn’t buy it. They thought that color-naming was universal, that all languages were on the same path toward complete knowledge of color names, and that the only difference among languages was how far along the path they were.

So the hypothesis was that maybe something about those colors, or the human brain’s perception of those colors, is more fundamental than culture—like, the human brain has a “deep grammar” for color, a hard-coded categorization of the color space.

“There was, as you’d imagine, massive pushback,” says Delwin Lindsey, a biopsychologist at Ohio State University. Despite the bouquet of languages spoken by the subjects of the study, they were all San Franciscans with similar backgrounds. Even in a town famous for embracing diversity, that’s not good enough for science.

So Berlin and Kay went big. They sent teams into the field, finding people who spoke 110 different languages and showing them, one by one, the chips in the Munsell array—adding 10 “achromatic” chips, neutral colors. The field workers were supposed to ask for “short names” for the colors of the chips. And then, once they had the gamut of basic color terms, the workers would show the subjects the whole palette, a board with all the chips. And the native speakers were supposed to point to the best example of each basic term. Those findings became the World Color Survey, still one of the best databases on how people talk about color.

The result: “The location of the focal colors, the best exemplars of the basic color terms, is in a remarkably similar place across different languages,” Conway says. And in fact, as cultures get more industrialized, they get more color words. First comes black and white (or light and dark), then red. Then a bunch of others.

But why? Why should humans all choose roughly the same places to identify transitions, when color is just photons, irreducible quanta of light on a continuous spectrum of wavelengths? Linguists might say it’s because language creates cultural norms. Anthropologists might say it’s because some colors have more cultural relevance than others. Neuroscientists and physiologists might say it’s because of the specific light-sensitive cells in the primate eye tuned to pick up red, green, and blue wavelengths and send signals to the visual cortex—trichromatic vision. “It’s a major dividing line across anthropology, psychology, linguistics. This research is massively cross-discipline. Everybody’s interested in this,” Lindsey says. “What are the origins of thought? How do humans build mental representations of the world?”

Berlin and Kay’s answer, by the way—and currently the dominant theory for all this—was that some colors are more “salient” to people than others—more important, somehow, to how the visual system works.

For all its strengths, the World Color Survey also has weaknesses. The investigators asked the participants to define their color words first, and not every language had them, for one thing. And the methodological rigor of the field researchers varied pretty widely.

Conway and Gibson’s team set out to talk to the Tsimane’ with an improved set of methods. They also included in their study group native Spanish speakers in a nearby town in Bolivia and English speakers in North America. And they took the Munsell chips.

But this time they played a game. “We asked people to describe color chips in a way that would make sense to someone else who spoke their language,” Conway says. “Instead of saying, ‘You have to use a basic color term’—and telling them what that is—we just asked people to describe these chips with whatever terms they thought were useful.”

That’s where Gibson’s information theory expertise comes in. The team did a statistical analysis to figure out the “modal term”—the color word—for each chip in each language.

“Then you just go through for every chip how effective people are at communicating those labels,” Conway says. “If you have the complete color array sitting in front of you and I have it in front me, and I pick a chip without showing it to you and use a word to describe that chip, how many guesses does it take you to home in on which chip I’m talking about?” Some colors are highly communicative; it takes fewer words and guesses. Some take more. In information-theoretic terms, some chips require more bits.

In general, North American English and Bolivian Spanish are pretty sophisticated about communicating color. Tsimane’, as Gibson had explained to Conway at their first lunch together, doesn’t have as many color words and doesn’t use them as precisely. “But the kind of surprising conclusion was that these three languages, even though they’re totally different, have pretty much the same pattern when you rank-order the chips by their communicative efficiency,” Conway says.

In other words: An English speaker and a Tsimane’ speaker have roughly the same difficulty conveying the same colors.

Munsell color chips rank-ordered left-to-right by how effectively the color chip is communicated. Each row shows data for a given language. The rows are ordered according to the overall communication efficiency of each language.

Richard Futrell

Gibson and Conway’s team went back to the WCS database and did the same math. It held. No matter what language you speak or what culture you come from, some colors are easier to talk about than others. If you make a chart with all the WCS languages arranged from lowest sophistication to highest on the y-axis and colors in order of easiest to hardest on the x-axis … well, you can see what happens.

“No matter how sophisticated, all languages in the world communicate more effectively about warm colors than they do about cool colors,” Conway says. “That’s none of the Western ideas about color. That’s a much more primitive, fundamental, backbone universal color categorization.”

But, again, why would “warm” colors be easier, in a sense, for a language to acquire a way of talking about? Even Conway, an artist himself, was surprised. “This is the main, first-order terminology we use to talk about pigments and their importance in painting,” he says. “But painters are sort of pooh-poohed in science.”

OK, so, new experiment. Conway and Gibson acquired another database: 20,000 images Microsoft uses to train machine learning systems. The pictures were of random objects, hand-coded, pixel-by-pixel, as to what was the object and what the background was. The human brain is very good at discerning the difference.

The researchers then did a statistical analysis on the colors of those pixels. Your intuition probably tells you that the colors in any given picture of the world should be arbitrary, right? “We are so tuned to color, so our experience of the world makes it seem like colors are radically different in one place than another,” Conway says. But they ain’t. People have done the math. “The color statistics of the natural world are virtually identical, regardless of biome and ecosystem.”

Once you apply the object/background distinction, though, things come into a very different focus. “Objects are systematically biased for—drumroll—warm colors,” Conway says.

It’s weird, but not unacceptably so. Hypothetically, the green of a forest or the blue of sky or ocean are so vast they become, in a way, less apparent. They’re big; they’re background. “The cool colors are in the background. Greens and blues are typically ones we don’t want to label. These are not objects,” Gibson says. “Warms are the humans and other animals, the berries and the fruits, flowers and stuff. All these things tend to be warm-colored.”

As a hypothesis, this idea provides not only an explanation for the array of languages and how well they convey warm-versus-cool but also for why languages acquire new color words in a predictable order: usefulness. The colors that are more a part of daily life enter the lexicon sooner. That’s a very different way of thinking about color than Sapir-Whorf, or anthropologically, or neurophysiologically.

While colors don’t vary much from biome to biome, the colors of the built environment do. Nobody would confuse Hong Kong’s lite-brite nighttime cityscape with a crystalline autumn morning in Boston. “One of the great advances of industrialization is the expansion of the gamut. We’re now at the point where we have HD color televisions capable of producing many more saturated colors than we’ve ever been able to produce,” Conway says. “The earliest cave artists had two kinds of ochres and charcoal. This radical change to our environments that we brought about is the addition of synthetic pigments.”

People in uncontacted tribes can see the same colors as someone who grew up in Paris or Tokyo—sunsets are just as beautiful and have all the same variations. But it’s the advent of new pigments and dyes that give objects arbitrary colors beyond the warm/cool split. Languages get industrialized, and they need new labels for the new colors.

Now it’ll be up to the rest of the color vision community to confirm or overturn the idea. “It’s the right way, to me, to look at it. It’s not the only way, but it’s better than basing it on an ethnocentric notion of primary colors and trying to see some kind of progression,” says Qasim Zaidi, of the Graduate Center for Vision Research at the State University of New York. “I think it’s the most interesting and useful paper on color names that I’ve read.”

Early work with the Tsimane’ tends to support the hypothesis. Play a color game with native speakers of English, say: Show them an object, ask them to describe it, and then show them the same object in another color. Do it with four natural objects, and then four artificial objects. Like, start with a green bell pepper and then a red bell pepper. More often than not, they’ll describe the second object in terms of its color, spontaneously.

  • More on Color

  • Adam Rogers

    The Science of Why No One Agrees on the Color of This Dress

  • Adam Rogers

    Art Fight! The Pinkest Pink Versus the Blackest Black

  • Adam Rogers

    A Year Ago, The Dress Murdered the Idea of Objective Color

Not the Tsimane’. Generally, when you show them a yellow banana and then a green one, they’re more likely to describe the second one as “unripe.” (Whatever cognitive energy they don’t put toward color they do use on botany, with a knowledge of thousands of plant species.) But switch to the artificial objects, and on the second cup or whatever, they switch to color descriptors. “Color has been isolated through industrialization as the only way of discriminating those objects,” Conway says. “You’ve promoted the usefulness of it.”

It won’t be easy to test the hypothesis; people study color not just because it’s a fascinating way to get into someone else’s head—is your red the same as my red?—but because it digs into how the machinery of human perception and cognition. Somehow, with nothing more than eyes, retinas, and a hefty tangle of neurons, people build entire, colorful worlds in their heads. “We just don’t understand enough about how high-level color vision is done by the brain,” Lindsay says. What we talk about when we talk about color is nothing less than the human brain making the cognitive world.