Perception and primary colours

I understand that the choice of RGB and CMY as primary colours has a lot to do with our perception and visual apparatus, rather than fundamental physical principles, but…

Given that (on my computer screen):
Cyan can be made by mixing equal proportions of green and blue
Magenta can be made by mixing equal proportions of blue and red
Yellow can be made by mixing equal proportions of red and green

Why is it that (to me, subjectively, at least):
Cyan appears a ‘closer match’ to blue than to green
Yellow appears ‘closer match’ to green than to red
Magenta appears ‘closer match’ to red than to blue

Or, to look at it another way
Cyan appears a ‘better match’ to blue than does magenta
Yellow appears ‘better match’ to green than does cyan
Magenta appears ‘better match’ to red than does yellow


Going out on a limb, I’d say that it is a personal perception issue. I imagine that half the people would view true Cyan (not the colour that Windows calls Cyan) as closer to blue and the other half as closer to green.

Ditto for the others.

Just my opinion.

I argued with a friend on this point once. We couldn’t figure out whether a particular color was blue or green. I said it was blue, he insisted that it was green. Turns out that it was cyan, so we were both right.

Whether it’s what you would consider true Cyan, the color that Windows calls CYAN is a true and equal mix of BLUE and GREEN, unless there’s something wrong with your monitor.

Cyan is more green than blue to me, for what it’s worth.

The physical output of the R/G/B cones is further processed in the retina in such a way that the color information is represented by two channels which encode between opposite values of red/green, and yellow/blue, respectively.

Think of this transformation as if the familiar compass directions of N/S and E/W get projected onto a new coordinate system whose axes are not at right angle, like NNE/SSW and NW/SE. It’s perfectly possible to describe a color plane on such axes, but the lengths are off and you can’t change one value without affecting the other. You can imagine what this does to the perceived distances and directions on Newton’s color circle.

This opponent processing model also explains why you can’t perceive a “reddish green”, despite there being more or less independent red and green receptors in the eye, and why the after-image of a red object is green, not cyan, as one would expect from a pure RGB model.

Also, separate from the issue of primary colors and whatnot is the idea that there are eleven “basic” colors in English: white, black, red, green, yellow, blue, brown, purple, pink, orange, and gray. (Cecil on this subject.) What this means is that we* think of every color as a shade of one of these eleven. So cyan has to be something, and I think that most people would say blue, although I know some say green. I doubt it’s half and half, as Achilles speculates.

    • When I say “we” think this way, I mean normal people. Every time I say this, someone will respond with “I don’t think of magenta as a shade of red!” Use this test: If someone points to a magenta bird and says, “See that red bird?”, do you say, “Where?”

While we’re on the subject, here’s something that’s been bothering me for a while: If the primary colors for subtractive (pigment) mixing are magenta, yellow, and cyan, which work just fine and dandy for printing, why do we use red, yellow, and blue as primary colors in painting? Why not straight M, Y, and C?

(from a minute or two search)
“The Primary colors, red, blue and yellow are colors that are not created from any other combination of colors.”

I don’t know how much I trust it, but I’m about to go to sleep. I’ll research a bit more tomorrow.

Another possible explanation is that the human eye is most sensitive to green, then red, then blue. Yellow, being lighter than either red or green, looks more like green because green looks lighter. With magenta, the difference is stronger because the eye’s blue-sensitive cones don’t register as brightness.

However, the colors on a computer monitor or TV aren’t ideal red, green, and blue. They’re not even close. The green resembles yellow because it is itself yellowish; ideal green is closer to the hue of a green traffic light. Monitors don’t use ideal green because mixing it with red would not produce a very saturated yellow.

In theory, all the colors available to the human eye can be made by stimulating the 3 types of cone cell in varying amounts, but in practice the spectral response curves overlap too much, so that there is no wavelength that stimulates only one type of cell. Cite.

In theory, we can surmise the ideal colors by measuring the invariant hues. Invariant hues are wavelengths that do not appear to change color when their brightness is increased. As you can see by the chart in the cite, red light stimulates both red and green cones, and blue light stimulates both green and blue cones. When a very bright red light is observed, it appears to change hue (becoming more yellowish) because red cones reach a ceiling of maximum stimulation, allowing green cones to catch up, so to speak. Therefore red is not an invariant hue.

There are 3 invariant hues, and they have been measured at 574, 507, and 476 nanometers wavelength (sorry, no cite as I got this out of a scientific article a couple years ago) which we perceive to be yellow, teal green, and azure blue, respectively. The yellow wavelength corresponds to where the red and green cones’ spectral response curves meet, so that they respond equally to that wavelength; when the light intensity increases they reach the “ceiling” simultaneously and the hue doesn’t change. The green hue corresponds to the wavelength where the red and blue cone response curves meet; this wavelength does not shift towards either bluish or yellowish when intensity increases. The blue invariant hue, however, corresponds to where the green and blue cone curves cross. Therefore the true primary color “blue” is not really blue. It is more of a deep indigo/purple.

True red, green, and blue are not the same as the colors displayed by individual phosphors on your monitor, and true yellow, magenta, and cyan do not resemble one primary color more than the other.

It is a common misconception that red, yellow, and blue are the primary colors. In the cases where these 3 colors are used successfully, the red is typically more of a magenta and the blue is universally a shade of cyan.

>>“The Primary colors, red, blue and yellow are colors that are not created from any other combination of colors.”
Well, this is not very helpful teaching. Trouble is, human eyes are very good at making us aware of object’s properties in ways that suit our evolved activities, without making obvious to us the mechanisms involved.

We seem to get color information from the interaction of three different kinds of light sensors, with their own sensitivity curves that overlap each other. Only by many ingenious experiments, including experiments with people who are color blind, have folks pieced together what’s going on in there. And they did come up with three primary colors that are useful, of which all observed colors are some combination - but the primary colors are imaginary. These primary colors are more saturated than pure spectral (monochromatic) colors.

Now, if you want to have the best possible gamut (color range) from mixing three light sources, you should pick a green of about 520 nm, a red of at least 600 or 630 nm or so, and a blue of less than 420 or even 400 nm. Pretty near everybody is going to look at those and call them green, red, and blue, though the blue is debatably violet or indigo. Definitely no yellow. You get excellent yellows mixing the above green and red. But you can have a larger gamut if you include more light sources, especially a second green of maybe 500 or 510 nm, and then a second blue of maybe 430 or 450.

I disagree about the red and green. 630nm is too short a wavelength, and 600nm is plain orange. There is no way a system could imitate, say, the color of a 670nm wavelength by mixing 630 and 400 nm; the resulting color would have too much lightness and not enough saturation. The best red would be as deep a red as possible, which 670 would accomplish but something like 700nm would be preferable.

Mixing 520nm with either 420 or 400, while still producing a better cyan than a computer monitor can, still won’t produce as saturated a cyan as, say, a pure 480nm. But mixing a 670nm wavelength with 520 won’t produce a very saturated yellow. With the wavelength of the green channel on a 3-channel system, too much is not enough.

To get the maximum practical range of colors, a system would need not 3 color channels but 4, around 670-700, 580, 505, and 400-420 nm. Not only would this produce a full range of hues with suberb saturation, but if the 4 channels were independently controlled it would provide for mesopic light conditions (e.g. reading the Dope indoors at night on the computer) where rods contribute their own spectral sensitivity, causing the human eye to function tetrachromatically rather than trichromatically.