Red, yellow, and blue are NOT the primary colors


I was inspired by Liz Elliot’s Magenta Ain’t a Color (via) article to write out my favorite color theory rant: art teachers are lying when they say that red, yellow, and blue are “the primary colors.”

There are two ways that color is created in the real world: in the additive model, used by light-based devices such as your computer monitor, uses red, green, and blue. When all three are combined in full strength, they create white in other combination, they create all the other colors you see on your screen.

The subtractive model comes into play when physical pigments are mixed. The primary colors there are the slightly less familiar cyan, magenta, and yellow. Again, they can be combined to create all the color you see on the printed page of a magazine (actually, the printing process also uses black, because combining the three dyes makes a muddy, unsatisfying black).

What makes these color models so obviously correct is how they interact — combining any two colors from one model creates one of the primary colors of the other model! Check out my graphic above, or open your favorite graphics program and create one for yourself.

So what the heck is “red, yellow, and blue”? It’s a lie, plain and simple. It started when people didn’t know any better, and it’s perpetuated by art teachers everywhere who for some reason believe that it makes color easier to understand than the truth.

It’s a shame, too, since the RGB/CMY color models are strikingly obvious and beautiful, but a little tricky to understand and remember if in the back of your mind you’re tending to revert to RYB. If art teachers could get with the program, we’d have a much easier time understanding how color actually works.

14 thoughts on “Red, yellow, and blue are NOT the primary colors

  1. Well, no. CYM works great when the subtractive colors in question are transparent, like dyes and printing inks. But with opaque pigments, such as those in oil or acrylic paint, RYB is a lot closer to reality. You’re going to find it impossible to mix that secondary ultramarine in the CYM diagram with cyan and magenta paint.

  2. I’m very glad you cast some light (no pun meant) on this confusion, perpetuated, as you say, by imbecile arts teachers nationwide. Perhaps you’ll do the same with this blasphemous “evolution” theory that science teachers use to poison our children’s minds next. Thanks, and God bless.

  3. Franklin~

    You know, I almost e-mailed you to ask you to comment to get your perspective. The paint mixing point has been brought up to me before. But it seems to me that the weakness of the RYB model as a tool for mixing paints is revealed by this example: Mix equal amounts of red and yellow paint, and you get a slightly orangish red, very close to the original red. Mix equal amounts of yellow and blue, and you get green, a completely new color.

  4. That’s because pigments have different tinctoral strengths, so equal quantities of them don’t produce exact midpoints of hue. A lot of thought had to be put into printing inks to stop them from doing the same thing.

  5. I thought about this some more on the drive home today, and I want to amend my last comment. While you wouldn’t want to use tinctorial outliers, you don’t have to worry as much about tinctorial strength in printing ink because you’re not mixing them together – you’re layering them on top of one another. Equal physical mixtures of CMY primaries might also fail to deliver perfect secondaries (I would predict as much), but that wouldn’t blow the CMY model.

  6. As I’d expected, you’ve gone a bit over my head. But it seems to me that all the effort that went into the CMY “printing inks” would lend them credence as being primary. In contrast, here’s what we get when we mix RYB:

    The lack of difference between red/yellow and their mixture and between blue/yellow and their mixture seems to point to a weakness in this system — you’re getting much more coloristic umph from one combination then from the other. Doesn’t that suggest something fishy?

    I’m not saying that RYB is completely without merit; but of course mixing any colors begats other colors. It seems that the onus is on RYB to prove its primariness in light of the evidence to the contrary.

    I suspect that there is something to your distinction between mixing pigments and layering them on top of each other. I also am not aware of commonly available cyan and magenta artist paints. But I think the real test would be an artist brought up — from birth — learning CMY/RGB, and compare them with RYB artists in tests of matching particular colors and in general their ability to render color. Is it possible that mixing CMY paints would turn out to be more effective, after all these millennia?

  7. But neither the CMYK nor the RGB result colors are really “combined” in the same sense as mixing colors; they are laid in patterns that sometimes overlap but not necessarily (offset screens in printing, beams of colored light in electronic screens). It’s the eye that makes the combined color by proximity.

    Mixing printing inks like you do with oil or acrylic paints produces very crappy results.

    I think the artists you are looking for are Seurat, Signac and Chuck Close.

  8. The model falls out of working with the materials, not acculturation. Golden Paints makes a Primary Magenta and a Primary Cyan. Pick some up and see if you can make secondary blue by mixing them. Get some Cobalt Blue or Ultramarine for comparison. I think you’re going to end up with neutral purple, but I haven’t run the experiment.

    What evidence do you have against RYB as primaries? Have you ever tried to mix red, yellow, or blue from other hues of paint? Meanwhile, you can easily mix RYB secondaries from the primaries.

    RYB has one annoying weakness – you can easily find colors, and even buy paints, that go way outside of the RYB gamut. Notably, magenta. But I’ve run into that problem with RGB too. The color of the sky in the original art for this comic was sherbet orange. RGB supplies a terribly small space in the gamut for orange, so the scan turned out crazy, and I soon found that I could either balance for the orange, or balance for everything else, but not both without resorting to unacceptably complicated selection. Even then I wasn’t getting that creamy orange that I had made so easily with gouache out of cadmium orange and zinc white.

    Typically the weakness of the RYB gamut doesn’t get in the way when you’re painting. Especially in figurative painting, usually you’re trying to modulate intensity, not maximize it. (Colors come out of the tube already maximally intense.) Here RYB really shines, because you modulate intensity by mixing complements, and the RYB complementary pairs (RG, OB, YP) foil each other almost perfectly. Between that and the ease with which you can mix secondaries from primaries, RYB supplies a great model for painting. If you need to go off-gamut, you just lay down the paint and see what happens. RYB doesn’t have a big gamut, but the gamut of pigments themselves dwarfs all three models and every other taxonomy besides.

    A thorough education in color would teach painting and RYB, printmaking and CYM, and photography and RGB.

  9. Of course, printers likewise go off the CYMK model all the time, in the form of spot color. I wish you could do that with a computer monitor.

  10. For painting, I just use the Munsell primaries-red, yellow, green, blue, and purple. Much easier and very effective way to get precise complements.

  11. The evidence against RYB primaries is exactly what I’ve got in the post — it doesn’t correspond to the way that coldhearted machines make colors in the real world.

    (Of course all the systems have a finite color gamut, though magenta is a curious color to not be able to create; neither of the other two systems exclude a primary color from another system. (Oh, but I forgot: magenta isn’t a real color!) As for your orange sky, perhaps the problem isn’t with the RGB, but with your scanner’s gamut. I cooked up some oranges:

    OR: Another possibility I just thought of is that the reason you’re perceiving orange as having a small gamut is because you’re thinking of “orange” as the range between two primary colors. Note that under both CMY and RGB, orange is the range between a primary color and a secondary color. It would be like if we had a separate name for the range between cyan and green, and you wanted a specific shade of that color. Right, right??)

    I particularly appreciate the observation that painters in practice rarely actually use red, yellow, and blue pigments as their sole color-creating source. You start with whatever base colors make sense for what you want to paint, and the RYB model is sort of a mental aid in getting to where you want to be(?).

    But here’s what it comes down to: if in fact RYB is corresponding to something in the real world, and not to something in artists’ heads, and it is in fact the result of “working with the materials, not acculturation” (which sounds like a chicken/egg problem to me), then why does it borrow one color from CMY and two from RGB? How do you account for the on-the-face incongruity of that?

  12. You cant really tint RGB colors, but try picking the value up on the Y and R ends of that gradient and you’ll notice the midpoints start to go to a pastel range that doesn’t look like the product of mixing white and orange in gouache, which keeps its luminosity and, dare I say it, personality.

    I’m perceiving RGB as having a small space for orange in the gamut because I’m looking at the right chart.

    Cold-hearted machines that mix housepaint to spec use a Hue-Saturation-Intensity model, not CMY. Does that mean that CMY is less true, somehow? No, it means that CMY is an inferior system for mixing housepaint. I’m sure that HSI isn’t a good system for commercial printing.

    I don’t know what you mean by RYB “borrowing” colors from the other wheels. I suspect that when faced with the problem of reproducing images using dot screens, printers figured out that the RYB primary blue was too cool and the primary red was too warm. So they adjusted the primaries to suit the purposes of printing. The only real world those primaries correlate to is the real world of printing, and as Pantone will happily tell you, there is a lot more out there than what is represented by CMY. Why do you accept the idea of different color wheels for additive and subtractive color, but not transparent and opaque pigments? It’s the same issue – different things work differently.

    And pace Ms. Elliot, magenta is a real color.

    It’s true that painters don’t seek out primary red, yellow, and blue for their palettes. That would be a silly way to make paints for artistic use. Instead we have chemical compounds like sulfured aluminum (ultramarine) and cadmium, which can be prepared any number of ways to produce hues ranging from red to green. So the manufacturers try to make the most luminous, stable paints out of the compounds and the artists build a palette accordingly. Typically they’ll pick a warm and a cool red, a yellow, a warm and a cool blue, black and white, and then fill out with secondaries and earth colors. You can’t mix two primaries to make a secondary as intense as a chemical that is secondary at the molecular level. The earth colors, well, they’re just beautiful in their way. Then you mix them with one another with RYB in mind and it all works rather nicely. Give it a try sometime.

  13. I mostly skimmed the comments because I have to dash out of the house in a minute, but Franklin and Alex are on the right track (except for Alex’s thought that you get crappy results when you mix printing inks—they mix totally differently than paint, and once one has a facility for it one can get beautiful results!).

    The main issue I have with your model Alesh, is that you’re using the simulated computer mixing as if it were actual mixing of pigmented color: it’s not. The eye, layering and proximity does the work. We can simulate it with the computer, but similarly, printing techniques are designed for reproducibility, which is to say that one can print and achieve effects that the computer can NOT adequately simulate, because those effects rely on the eye. You’re comparing apples and oranges, just apples and oranges which happen to look exactly alike until you try to eat them (or mix them, or physically print them). It’s just not commercially useful to exploit the less well-known aspects of the interactions of printing inks because it’s a more technically rare skill to be able to predict and repeat. Basically, it’s an art.

    I’ve been having a lot of trouble lately documenting my prints for pretty much this exact same reason. Any digital image flattens perception, and there’s something that happens when looking at one of my prints in real life that never happens on the computer screen—there’s a tension in the layers of color which doesn’t translate.

    Okay, rambly, typing-fast rant over and out!  🙂

Comments are closed.