Why does purple = red + blue?

Strange-sounding question, but here’s the long form - hopefully I’m making sense:

What we perceive as color is light reflecting at different wavelengths. The lower end of the visible scale is red, the higher end of the visible scale is purple, the other colors are in-between.

Now, the wavelengths that we perceive as orange are between red and yellow, so it makes sense to me that if we were to mix red and yellow pigments, the result would average out to an orange color. Ditto for green, the green wavelengths are between yellow and blue, so it makes sense that this combination of pigments will yield green.

But purple is higher than blue. The spectrum isn’t circular that it should come “back around” to resemble a blend with red, at the low end of the scale. If my understanding of the pigments above is right (no doubt it isn’t, and someone will explain to me why), red and blue should average out to wavelengths in the yellow range, shouldn’t it? Why, then, will blue mixed with red yield a purple color?

There are two kinds of color combinations, additive and subtractive. When dealing with sources of light, color is additive, when dealing with reflective pigments, it’s subtractive.

Sunlight is white because it adds all the different colors of light together. The primary colors of light are red, green and blue, which is why computer monitors and TVs use those color pixels.

In contrast, if you mix every color pigment together you’ll get black[sup]1[/sup], because it will reflect no light. The primary colors of pigment are yellow, cyan, and magenta, which is why color printers use those colors of ink. In subtractive color, red + blue = purple.

[sub]1. Assuming theoretical pigments. In real life, you get the color of poopy.[/sub]

Good question–Some earlier threads on this, from March 2004, November 2002, and one from way back in December 2000 started by Yours Truly.

Your basic mistake was in thinking there’s any “averaging” going on. The reason, for instance, that yellow and blue makes green is that a typical yellow ink or paint reflects a spread of colours between orange and green, and adding blue (which will cover a spread from green to violet) leaves only green in common. If both were “true” pigments, reflecting only yellow and only blue, you’d see nothing, because they would have no wavelengths in common.

My understanding was always that monochromatic purple quite simply doesn’t exist – while other colour mixtures have an equivalent single-frequency colour, there are some that can’t be reduced that way. (White being another, more commonly known example.)
This wikipedia article contains a chromaticity plot of the human colour space, i.e. a two-dimensional projection of the three-dimensional colour distribution for a fixed brightness – the outer, horseshoe-shaped line are the monochromatic colours, those that correspond to the electromagnetic spectrum; they only represent the boundary of the gamut of human vision. Note that purple/pink hues lie between the edges of the horseshoe, i.e. violet and red.

I don’t think that is true, if I understand correctly what you are trying to say.

Confusion exists between pure monochromatic light and combinations because of the way colour receptors work. As is commonly known there are three types of receptors responsible for colour vision in the human eye. These receptors differ in their response to different wavelength, peaking at red, green and blue respectively. The cut-off for each type is not sharp however, so there is a great deal of overlap between the different types. The result is that a monochromatic green light will stimulate both blue and yellow receptors, though not equally depending on the exact wavelength, and will be indistinguishable from a mixture of pure yellow and blue light whose net stimulation will be the same.

The key is that every visible light wavelength creates a unique pattern in our brain, like a three-value RGB code, but actually more complex. This is illustrated by another graph in Half Man’s cite: the color matching functions. In that graph, you can see that the three curves overlap by different amounts in a unique way at every point in the visible spectrum.

In particular, note that red has two peaks, the main one in the “red” end of the spectrum, and a smaller one in the “blue” end. That is where purple comes from.

Of course, monochromatic purple does exist, just like any other wavelength of light (visible or not). Unlike any other color, purple tickles the red and blue receptors a little, and green not at all. The brain interprets that stimulus as purple.

It’s a fact that a certain combination of red and blue light (with no green) looks like purple light, but that might as well be an accident. It’s an artifact of the eye’s physical and chemical make-up. All of the color-combining rules that we learn in school might be entirely different if our eyes were different. For example, purple might not exist (as a color) if our “red” sensors only detected red wavelengths.

Maybe there are some color-blind dopers out there who wouldn’t mind telling us how much sense there really is to rules like “red plus blue makes purple” or “blue and yellow makes green.”

Malacandra is talking about pigment: color by subtraction. If a paint that absorbs all wavelengths except yellow was mixed with a paint that absorbed all wavelengths except blue, the result would be black.

Yes. Also, ticker, what are these *yellow *receptors of which you speak? :slight_smile:

Actually, I believe it’s more like the other way around; those colour matching functions aren’t the actual responsivities of the cone cells, but basically mathematical constructs used to obtain a colour distribution equal to that of a standardized observer. The actual response functions can be found here; note that red doesn’t have two peaks.

No, there isn’t any single frequency of purple light. You may be thinking of violet, which is light of any frequency below blue, but true purple, and likewise pink, magenta etc., is a non-spectral colour. There’s no purple in a rainbow, in other words.

(Nit: I think you mean above, not below.)

Sorry, I hadn’t considered the distinction between purple and violet. I always thought of purple as “dark violet,” a low-luminosity version of the violet I see in rainbows.

I still think there’s some merit in the observation that red sensitivity spans the visible spectrum. The “one peak red” graph you link to does not plot the red or blue curves all the way to zero, as it does for green. I believe this is misleading–where does the red curve actually end?

Even if the actual red sensitivity does not peak again in the violet range, it must be there, and indeed it must have a significant perceptive effect. Otherwise, why does the mathematical construct you dismiss so readily require two red peaks to reproduce the sensation of violet light?

The OP’s question applies as much to violet as to purple, and violet definitely is a monochromatic hue. Computer monitors do not emit violet light, any more than they emit purple. But they do mimic violet by mixing red and blue. How is this possible, if violet light doesn’t tickle both the red and blue sensors in our eyes?

So, to dodge the violet/purple confusion that got me, let’s restate the OP question thus: What the heck is Magenta? I’m willing to accept (without the certainty of testing it for myself) that there is no single, pure wavelength of light that produces the same sickly sensation as, say, a Fuchsia divan. But nothing in this thread so far really explains how a mixture of red and blue light produces that sensation.

We’ve established that the brain can extrapolate color. But how? Is Fuchsia the color we’d see if we could see ultraviolet, or infrared? Or is the mechanism similar to how taste buds respond to, say, strawberry-bannana chewing gum? And how does that work? Is there a “flavor spectrum,” perhaps multi-dimensional, that relates chemical structures to flavor responses? Should I go to bed now?

EEeew. Vomit flavor!

When light hits the cones of the eyes, the cones generate a signal based upon how well they each absorb that spectrum spread. Instead of sending those signals straight to the brain, a multiplexing of the signals first occurs. What winds up getting sent to the brain are three signals, but not the response of the S, M, and L cones. The resulting signals represent brightness, red-versus-green strength, and blue-versus-yellow strength.

Brightness is easy enough to understand, but the other two signals might not be obvious. I’ll try a text diagram.

   red      magenta
     /     \
    /       \
   /         \.blue
   \         /
    \       /

So this diagram takes out the brightness component from the equation, and maps the red-versus-green strength to the Y axis, and the blue-versus-yellow strength to the X axis. As to what magenta means in terms of these three signals, magenta sends signals of high brightness, high redness, and moderately high blueness.

Interestingly enough, purple would send a signal of moderate brightness, high redness, and moderately high blueness – essentially, a dark magenta.

One thing I do want to point out is that yes, the eye does have yellow color receptors. Despite the color shown for the L curve on that Wikipedia page, the peak for that curve is at a frequency that we would describe as yellow, not red. The M curve, similarly, peaks at a color that is yellowish-green, not a strong green. This information and more can be found here.

Err, yes. :o Whoops. I blame being used to the visible spectrum being generally ordered according to wavelength, with the violet/blue part to the left…

RGB computer monitors generally can’t “mimic” violet, precisely because by definition, the mixture of red and blue is “purple”. Purple (or magenta) is the closest we can get to violet without actually producing violet light. Fortunately it’s close enough for most purposes.

The question becomes, then, why is purple close to violet in our perception? Violet should stimulate the blue receptors highly and the red/green receptors not at all, while purple should act as Punoqllads’ diagram shows.
Powers &8^]

It may not be circular, but it sure can be expressed that way.

Wait a minute. You aren’t saying an RGB monitor actually produces light of any other color, besides red, green, and blue, by mixing colors, are you? What you said could be construed to imply that violet is an exception in mixed colors. For example, since the word “yellow” applies to both mixed and spectral yellows, an RGB monitor can emit actual yellow light by mixing red and green. I doubt you believe that.

Instead, I think you’re saying the RGB spectrum somehow skips “rainbow” violet, producing instead this artificial “purple” that, by definition, can be achieved only by mixing red and blue.

In fact, Punoqllad’s Color Vision Wikipedia cite defines purple as a mixture of red and violet, which makes much more sense to me. I tend to agree that the answer to my question (and the OP’s, probably lies in that cite, but I need to think on’t.

You are correct that I don’t believe that.

Yes. Was I not clear?

The thing is that “yellow”, to use your example, can be mimicked by an RGB display, because the mixture of red and green looks is sensed by our eyes very similarly to the way pure yellow light is. In that respect, violet is an exception among mixed colors, in that it can’t be created by mixing colors at all! Only purple can be created from red and blue (or violet) light; violet can only be properly perceived as violet by emitting light of the appropriate wavelength.

Perhaps we were using different definitions of “mimic”? By “mimic” I was referring to yellow being mimicked by combining red and green light. In that definition, violet cannot be mimicked by an RGB display. If you meant “mimic” in that purple “mimics” (as in approximates) violet, then you’re right that computer monitors “mimic” violet.
Powers &8^]

Not exactly. It says “purples”, not “purple”. If you find the midpoint between spectral blue and spectral red, and find the midpoint between spectral violet and spectral red, you will find them both in the region of colors that would be described as purple. However, the latter would be more saturated; that is, perceptually further away from the color white.

There are lots of colors that can’t be created by mixing RGB primaries. Very saturated cyan tends to be the worst.

Yes, you were clear in what you said. I just imagined an incorrect inference that some folks might make from it, so I wanted to clarify what you were not saying.

I think I understand you, and I agree in theory. But I don’t expect a typical RGB display to reproduce any color very accurately, so I wonder if the difference between RGB “violet” and spectral violet is as significant as you seem to be saying.

Maybe RGB displays do a better job than I give them credit for. Someday I might do some experiments with a prism and a digicam, to see for myself how much visible spectrum an RGB representation can really mimic.