I'm taking a break from shopping for one of those swanky wide-gamut monitors (I'm looking really hard at the NEC MultiSync PA241W) to indulge in a little color science nerdery. For a long time the rather modest gamut of a standard monitor was good enough. It could easily reproduce the colors of common output devices and then some. Most serious work was printed to CMYK offset in the form of ads and magazines pages. As you can see in the above plot, the gamut of a generic CMKY profile fits easily within my monitor's gamut. But things changed when we started sending output to wider gamut devices like inkjet printers with eight inks and started capturing images with digital cameras capable of recording a very large gamut. Suddenly we find ourselves dealing with colors that can't be reproduced on our monitor. This is a problem for obvious reasons.

Have you ever wondered why monitors have limited gamuts in the first place? Why can't they be made to simply reproduce all colors? Where is the bottleneck? It turns out that this isn't just a limit in manufacturing, but is a theoretical limit of tristimulus (three color) models of color. That's a mouthful, but it's really quite simple. It all starts with our eyes: the human retina has three different kinds of cones cells which respond differently to different parts of the spectrum. The relative response of each of the cone cells looks like this:

When you see a color, lets say a pure 500nm color that should look bluish-green, you will get a small response from the ρ-cells a little more from β-cells and a pretty strong response from the γ-cells. The idea behind tristimulus color models is that if you could target each type of cell individually you could simulate the the effect of any color by varying the relative strengths of just three primary colors—essentially hacking your vision. You might pick primaries that correspond to the peak sensitivities of each cell, and when you want to get a response identical to 500nm light you would present a little red light to stimulate the ρ-cells, a little more blue light for the β-cells and a moderate amount of green light for the γ-cells. This is how RGB monitors and televisions work and it would work perfectly if you could find primary colors that could individually target the three types of cells. But nature threw a wrench in our design by allowing the response curves of the cells to overlap. The result is that with the exception of the extremes of the spectrum you can't target one type of cone cell without inadvertently stimulating one or both of the others—imaginatively called 'unwanted stimulations' in the color science literature. The effect of having an unwanted stimulation is a reduction in saturation of the stimulus. This problem doesn't completely break the model; the math becomes a little more complicated, but you can still simulate a lot of colors with just three primaries. You just can't reproduce them all—three color reproduction will always have a smaller gamut than the human eye despite the fact that the retina is in some sense a tristimulus device. You can't even specify them with a three color model without some slight of hand. A relatively wide gamut like AdobeRGB is still considerably smaller than the human eye. To create a space like CIELAB that can completely encompass human vision you need to resort to imaginary primaries—colors that exist as mathematical constructs but don't exist as real colors.

It's important to remember that this is a limitation of the model and not of the nature of color reproduction in general. One of the earliest methods of color photography, in fact, was capable of reproducing the entire visible gamut. Gabriel Lippmann won a noble prize for it in 1908: see Lippmann's and Gabor's Revolutionary Approach to Imaging.