Does color theory work the way it does (blend blue and yellow and you get purple) because of the nature of the color cones in the eye? How would color theory work if our cones were optimized for purple, orange, and teal instead of blue yellow red?
Color theory could choose three other colors as primary colors. Here is an excerpt from a posting from the alt.fan.cecil-adams newsgroup (remember newsgroups?) about 11 years ago:
Mmm. Several things seem misleading here.
Color theory doesn’t say blending blue and yellow gives you purple. Blending blue and yellow light gives you white, if you have the right mix. Blending light is the simplest way of considering these things, because it’s light that your eye detects, and because pigments appear however they do through a slightly complicated interaction between the pigment itself and the light you illuminate it with. Moreover, we have certain adaptations that help us notice pigments and ignore the lights they are illuminated with, if enough visual clues are around to help.
We are trichromatic, but, oddly, we actually have four receptors having different color responses. We have rod cells, too. But for some reason we don’t use them to differentiate colors. We only use them to provide sharper details in small areas we concentrate on, and to extend our vision into lower-light situations.
Whether there are an infinite number of primary colors depends on what you mean by “primary color”. With three different kinds of cells used to measure color, we can describe color with two degrees of freedom and assign a third degree of freedom to overall brightness. So, you can make a map on paper that completely covers “color space”, and by plotting three noncolinear points on that map you can define three primary colors, and you can describe all other colors by linear combinations of your three points. If you can only have positive amounts of the three points, and if the points are all visible combinations of wavelengths, you can mix the three to make any color inside the triangle they define. Such a triangle is called a “gamut” and is interesting in graphic arts and designing television screens and so on. If you can also have negative amounts of points you can also have imaginary colors, and it’s been most convenient in color science to do this with three imaginary colors X, Y and Z which are kind of like red, green and blue except they are more intense than pure spectral colors. That is, if you mix the right ratio of X and white you get pure red.
So, in the sense that any three points can be used to create a full human gamut if we allow imaginary colors and negative values, you can make any three wavelengths primaries. You can also make any three other points primaries, too - you don’t need them to be single wavelengths.
However, as a practical matter, there is a fair amount of separation between the three color cell types, and you can pick the right red and green and blue wavelengths and create a real physical gamut with them that covers most of the human color space. With the red this is very easy - the map of the human color space on paper looks kind of like a rounded triangle with the red corner being quite sharp and neat. The blue corner is pretty good, too, though it’s a bit rounded. The green corner is actually fairly rounded, so this makes it hardest to pick 3 real physical wavelengths, but 510 nm is pretty good. We actually pay a lot of attention to yellows, which are red green combinations, and you want a longer green wavelength to get onto the yellow side of that rounded corner, but it works well enough for color television and photographs to look real and pleasing to us. A further challenge is that there are not many convenient physical light sources that are powerful and spectrally pure around 510, but if you spend a little more money it’s all doable. In practice, what we often lose is the ability to make some beautiful blue greens, but they don’t turn up often enough in real objects that it’s a big problem being unable to capture them.
So, in that sense, there are special primaries that are red as long in wavelength as you can see it, say 630 or 650 nm or so, and blue like maybe 400 nm, and a green that does require a little judgement but 510 nm works well.
A web search on “cie” and “chromaticity” turns up all kinds of references that talk about what I called “color maps” but in the context of color science are generally called “chromaticity diagrams”, and about color spaces.
Here’s just one:
http://hyperphysics.phy-astr.gsu.edu/hbase/vision/cie.html
“Blue and Yellow give you Green.” :smack:
Thanks for the replies.
>Blue and Yellow give you Green
If you’re mixing pigments, yes. If you’re mixing lights, no. Your eye figures out ratios between the amounts of light having different spectral distributions. Pigments are generally much harder to understand unless you take a numerical approach, multiplying reflectivity and source curves wavelength by wavelength and then integrating the product.
I’m an engineer that has to peripherally deal with illumination and color issues quite a bit…I attended a research seminar a couple of years ago by some LED research company (can’t recall which one) that showed that there is a small minority of the population that has 4-color receptor perception (either their rod cells could contribute additional color information to the brain, or they had a cone cell mutation).
They also showed charts showing that many genetic demographics have differing normalized peaks for some colors (e.g. peak perception of “red” can differ across demographics). The end result means that color sensitivity and perception across humans is far from uniform.
I also work with a “color standards” group of engineers that require a very stringent color-perception test to qualify for, and which I’m told less than 5% of people can actually meet (analogous to wine tasters and perfume sniffers). I found it fascinating that such a thing even existed 
Sorry for the lack of cites, but Wikipedia for “Color Vision”, “Trichomacy”, and “Tetrachromacy” has good info.
>color sensitivity and perception across humans is far from uniform
There’s also a trend over one’s lifetime. The transparent parts of the eye gradually become more opaque to the shorter wavelengths, first in the ultraviolet and later on for blue. These parts yellow, like old paper; lots of things gradually absorb more short wavelength light as time goes by, including us. So, babies can see well into the ultraviolet, and old people who get some of the transparent tissues replaced (I don’t remember if this is cornea replacement or cateract surgery or what) sometimes report their surprise that blue skies got much more brilliant blue.
About five years ago as I was taking my wife home from her first cataract surgery one of her first comments when we got outside was “Wow, I didn’t realize that shirt was so blue”!