Violet is not Purple: Is digital imaging broken?

Categories: Health, Rants and Rambling, and Science.

Last night, as I was trying to get to sleep, I had a series of thoughts that brought me to a troublesome gap in my understanding of human vision1. My problem? I couldn’t figure out how we see purple.

See, growing up, we were always taught that the spectral colors, starting from the longest wavelength, were red, orange, yellow, green, blue, indigo, and violet. If you ever asked the teacher what violet was, they’d say “it’s just another name for purple.” Indigo? “It’s kind of a puplish-blue. It’s named after a type of flower.” Thus, we are taught that the rainbow, and therefore the spectrum of colored light visible to humans, looks like this:

~400 nm                                                  ~650 nm
This Is Wrong

There is plenty of confusion to go around here, so let’s start with the basics.

Average human eyes have four types of receptors — rods and three kinds of cones.

The rods operate well in low-light, are far denser than cones, respond more slowly than cones, and help distinguish contrast and detail.

The cones require strong light, pick up colors, and help process motion more quickly. These cones each have a broad range of sensitivity. The short-wave ones, s-cones (often erroneously called “blue cones”) are most receptive around 420 nm. The medium-wave ones, m-cones (“green cones”), 534 nm; Long-wave cones, l-cones (“red cones” ), 564 nm. In reality, 420 nm appears violet to humans, while 564 nm is really more of a yellowish-green (and not anything like red at all).

Approximate Cone and Rod Responsiveness by Wavelength (normalized)

The responsivness of the l-cones is insignificant below 450 nm and above 700 nm; m-cones range from about 440 nm to 675 nm, and s-cones, from somewhere shorter than 400 nm to about 520 nm.

All that is sent to your brain (in terms of distinguing colors) is relative values from these three cones. So, consider a single, pure light wave with a wavelength of 570 nm. People with normal eyes will perceive this as yellow (because it stimulates both the m-cones and l-cones in the right ratio). Two simultaneous waves of light at, say, 510 nm (green) and 590 nm (red) in the right ratio will produce the same reaction from the cones — meaning it will appear to be the identical shade of yellow.

So, here’s where I got caught up: if purple (which at the time I thought was the same as violet) can be simulated as a mix of red and blue, how does violet — at the low end of the spectrum — stimulate the l-cones?

The answer is that apparently, it doesn’t. And the key to that answer is this: purple and violet are very different colors. Physiologically, violet results from stimulation of the s-cones without stimulating the m-cones. (Perceiving blue requires stimulation of the m-cones to some degree). Purple requires at least two wavelengths, so that the s-cones and l-cones are both stimulated without having too much stimulation of the m-cones. I’m going to have to grab a prism and play around with a few things to be fully comfortable with my understanding, but I think I have a functioning model again.

But here’s where things get odd.

The lowest wavelength that your screen can display is this:

Here is Blue

Exactly how that blue is produced (and its exact color) depends on whether you’re using a CRT or an LCD screen, and a wide variety of other factors. It’ll probably be around 460 nm, though. That’s right around what people like to call “blue.” Anyway, the fact is that your screen simply mixes Red, Green, and Blue together to make the colors that it can produce. And the blue that we see above is still stimulating your m-cones, or it would appear violet to you.

So, as far as I can tell, modern televisions, computer monitors, scanners — even digital cameras — simply ignore indigo and violet. There’s no way to record them, and no way to display them. Taking a digital picture of a violet flower or a bird with violet markings will produce an image that substitutes blue — probably dark blue — for violet.

A key example of this shortcoming is shown by any attempt to electronically render Yves Klein’s trademark “International Klein Blue,” which contains a lot of indigo and/or violet in it. In person, Klein’s art making use of this patented color is absolutely breathtaking, even if it’s just something simple like a sea sponge dipped in paint. Stunning. Unforgettable. Seeing a work like this in person is absolutely shocking:

You don’t understand how pretty this is

On your computer screen, it’s pretty unremarkable, isn’t it? To understand what you’re missing — and what’s wrong with stopping the spectrum at blue — make it a point to seek out some of Klein’s work the next time you’re near a modern art museum.

So, this raises an interesting question: why do all consumer electronics use only red, green, and blue? Consider that, if cameras and monitors instead used a red/green/violet color scale, we would be able to have the same range of color reproduction that we do currently, plus the visible colors from 400 nm to 460 nm. In practice, most people just don’t take much note of violet, and simply don’t miss it. But wouldn’t it still make more sense to be able to reproduce it when it is present?

I mean, doesn’t your monitor suddenly feel strangely defective now that you realize that there are colors you can see but which it completely lacks the ability to render?

1 The thoughts themselves started off with wondering whether a very low intensity LED (or similar light source) with a peak output around 496 nm — the peak sensitivity of the rods in human eyes — would be useful for assiting with seeing in low-light conditions without reducing night vision for objects not illuminated by the LED. In the final equasion, it seems that the use of red light works at least as well. But that’s the sort of random thought that goes through my mind when I’m drifting off.


  1. Cullen Jennings


    You have too much time on your hands but few extra things to through in your model …. Humans have 3 color receptors – you would think this would result in needing to describe color in a 3 dimensional axis plus of course the overall brightness – it does not. Somewhere in our brain, (not in the eye or optic nerve) these 3 axis get mapped to a perceptions with only has two degrees of freedom. Weird.

    CCD and CMOS sensors are sensitive into infrared band so manufactures need to put in place filters that block this and causes weird effects. The human eye allows almost no ultraviolet light to pass – actually some operations remove this filter which allows people to see wonderful colors on plants and such for about 2 weeks till the receptors burn up. Rat’s on the other hand can see into ultraviolet pretty well.

    There are lots of “colors” that are impossible to model on screen or standard printing – take any florescent color for example.

  2. A fine article. Your monitor approximates the sRGB colour space, which contains a mere 35% of the eye’s gamut. That’s equally shocking.

    Dammit, your captchas are hard!

  3. p e d morgan

    The eye’s red receptor has a secondary absorbtion/response at short wavelengths in the farther indigo overlapping the blue receptor. The brain interprets naturally this as the mix of red and blue i.e. violet in this case, and so it looks like purple to the brain.

    This is not easy to find on the web – I will post when I find the page again on a computer that does pdfs.

    Most pages about color are mistaken in this regard (as are essentially all school teachers and people that believe that the color circle is a physical reality – it is a brain reality!). I use the “question”? “Why does violet look like blue” in my university course on “innovation and creativity” as an example of the “unasked question” that leads to scientific progress when it is finally asked.

  4. p e d morgan

    Correction: I note in the above that I made a mistake “Why does violet look like blue” should have been “Why does violet look like purple”.