When working with images that contain a lot of foliage, I've noticed that what appears to me as green is often more yellow than green, at least according to the RGB numbers in photoshop. My eyes say green but the numbers say yellow. I've been noticing this for a long time (I've been using photoshop since version one). So one day I decided to get to the bottom of this—that was about ten years ago and in that time I've accomplished little other than coming up with a rather long list of questions about color perception. For instance, have you ever wondered why it's common in English to speak about a blue-green color, but you rarely if ever describe the color as bluish-red? A possible answer is that we use purple when we want to talk about bluish-red. It's common to talk about light-blue, but it much less common to hear someone speak of light-red because we call that pink. Our vocabularies contain color names for blue-green like turquoise and we'll occasionally use azure as a variant of blue, but they seem to belong to a different level in the color hierarchy than purple and pink. This is because in English purple and pink are what are known as basic color terms while azure and turquoise are not. If you are a native English speaker, you have eleven basic terms for color: black, white, red, green, yellow, blue, brown, orange, pink, purple and gray. This is not to say that you don't recognize or distinguish other colors, but the terms you use are generally alterations or combinations of the basic terms such as dark-blue and emerald green, or they are colors derived from concrete objects like salmon or robin's egg, or they are invented by paint companies—Oceanside (SW6496) from Sherwin Williams. If you are not a native English speaker there is a chance that your language breaks the spectrum up differently. Traffic signals in Japan, for instance, are more or less the same color as they are in America, but most Japanese will tell you that a blue (ao, 青) light means 'go'. Japanese also use ao to describe vegetables and foliage. As you read this, someone in Japan might be wondering why photoshop is telling them their ao foliage is yellow. The Japanese word for green (midori, 緑) is a relative newcomer to the language having arrived in the Heian period, and the boundaries between blue and green are not the same as they are for English speakers. Here's a fun Japanese color lesson. In the world of color categories the difference between Japanese and English are rather modest. Speakers of the Dani language in New Guinea famously use only two words for colors: 'mili' encompassing the black, blues, and greens and 'mola' for white, reds and yellows.

The Spectrum, it sure doesn't look continuos

When we look at the color spectrum or a rainbow we break it into categories despite the fact that the spectrum is a continuous, linear, smooth transition between wavelengths. There are no inherent qualities in the visible spectrum that define boundaries between colors, yet when we look at the spectrum we do two things: we perceive boundaries and we form names within those boundaries. It's important to note that these are two separate things: lexical color categorization and perceptual color categorization. If you are temped to think that without words for a color you might lack perception of the color you wouldn't be alone. William Gladstone, a pretty bright guy and a four-time Prime Minister of the United Kingdom, was of the opinion after his considerable study of the classics that ancient Greeks had underdeveloped color vision because they didn't distinguish them lexically.

Color, siting squarely at the intersection of language and biology, is fertile ground for linguistics, psychology, and biology. In 1969 Brent Berlin and Paul Kay published a landmark book Basic Color Terms: Their Universality and Evolution examining how different cultures perceive and name colors. Although their methods have been controversial (asking people from wildly varying cultures to identify Munsell color chips may not be as culturally neutral as you might think), their finding that a culture's basic color terms can be predicted by the number of terms in the language is probably sound. Cultures don't arbitrarily dice up the spectrum. If a language contains only two basic terms, they will distinguish light and dark. Adding a third term introduces red. Then green and yellow, then blue, and on from there until you reach the maximum of eleven. You won't find, according to their study, a culture with basic terms for light, dark, and purple or a language with that's carved the rainbow into scarlet, periwinkle, and olive. The basic colors around which we form linguistic categories appear to be universal and there seems to be a prototype effect surrounding the naming of colors. When a speaker of a language that has a term for red is asked to point out a color that best exemplifies the color red, the color he or she picks seems to be universal across languages. It would seem that we are hardwired to recognize certain primary colors which are universal despite varying widely on how we form categories around those colors. When asked to point out where the ideal red sits in the spectrum we more or less all pick the same color, but when asked where on the spectrum red becomes the next basic color, that's where we disagree. Interestingly when speakers of languages that have one word for blue and green are asked which color best exemplifies that color category, they don't pick a color somewhere in between, but instead pick the same focal blue or focal green which speakers of languages incorporating words for both colors pick.

It makes you wonder why.

In the late 60s, around the same time Paul Kay and Brent Berlin were publishing their work, Russell De Valois was studying color vision in monkeys whose visual system resembles ours and found a biological basis for the color model first proposed by Goethe and later molded into a proper theory by Ewald Hering called the color opponent process. The idea is that we form colors based on opposing hues: Red vs. Green, Blue vs. Yellow, and Dark vs. Light. This makes sense when you consider the way we think and talk about color. For instance, try to visualize a greenish-red. You can't really do it because in the opponent process red is an opposing force to green. The more green something is the less red it is—it's one or the other. Valois identified the cells in the eye that are responsible for identifying the opposing red/green, yellow/blue and black/white and the hues to which the cells respond most vigorously correspond with the focal primaries found in language studies giving a biological explanation for the prototype effects of color categories.

If you are a photoshop nerd you will recognize Hering's opposition theory as the underpinning of the LAB color space where colors are described in three channels corresponding to the opposite pairs green/magenta, blue/yellow, and light dark.

Alas, I still don't know why foliage that seems like is should be mostly green doesn't look right unless it's mostly yellow.

A few links:
The World Color Survey begun in the 70s by Berlin and Kay to help understand colors and language.
Some analysis of the data (with cool charts) is available here

Crayola Colors when I realized something was up with naming colors

Nathan Moroney, a serious color nerd at HP with a fun color zeitgeist application.