Another portable and accurate clinical test for colorblindness is the Farnsworth Munsell 100 Hue test, introduced in 1947. The test consists of 84 stubby cylinders, about the diameter of a nickel, with a color sample painted on one end. These surface color samples define equal, small steps around the hue circle of the Munsell Color System at a low chroma and moderate lightness (the colors appear to be muted, mid valued pastels). The hue circle is divided into four trays, and all colors within each tray have a constant lightness. Thus, only hue can be used to discriminate between the color samples in each tray. Subjects are asked to take all the color cylinders within a single tray and sort them left to right in a continuous hue sequence (for example, from red to yellow or turquoise to violet).

On the opposite end of each cylinder is a number from 1 to 85, indicating its correct place in the hue sequence around the Munsell hue circle. When the subject has finished sorting the cylinders, they are turned over and errors in the ordering are scored as the sum of the differences between the cylinder numbers and their numerical order in the sorted sequence.

When this task is given to "normal" individuals, most do not get a perfect score and mistake the ordering of some hue samples, but the errors are distributed randomly around the hue circle and involve hues next to each other in the sequence. When individuals with any of the three main types of color deficiencies take the test, they confuse widely different hues and the confusions tend to cluster in specific areas of the hue circle, as shown below.

color confusions and the three types of color deficiency
average confusion scores on the farnsworth-munsell 100 hue test (adapted from Kaiser & Boynton, 1996); approximate location of white points inserted; chroma and lightness have been enhanced for clarity and are limited by your color monitor gamut

Now if I follow the Chevreul [hue] circle, starting from red [at 6 o'clock], and going [clockwise] round towards blue [at 10 o'clock], in every division which I pass, the sensation of yellow becomes fainter and fainter ... until very soon [at violet] the yellow disappears altogether, and nothing but a dark grey or perfectly colourless hue remains ... the blue I see perfectly, but the various tints of violet are to me only a darkened blue. ... At about the second or third division beyond "bleu vert" [between blue green and green] the blue has entirely disappeared, and nothing is left but a neutral grey. Beyond this the illumination begins to increase again, and at the same time a sensation of yellow begins to enter; the light and the colour both gradually heightening as I advance, until at the division "jaune" [yellow, at 2 o'clock] the darkening influence has entirely disappeared, and the full normal yellow hue is obtained. (William Pole, 1859)

Most color deficient people remark on color discrimination problems among warm hues, less often among cool hues. The likely reason for this is not that these cool hue confusions do not occur, but that our color language is much more precise and discriminating for warm colors than for cool, and because purples and magentas are environmentally the rarest hues in both natural and artificial materials. As a result, color naming or recognition errors by "colorblinds" in the cool and "extraspectral" hues are harder (less common) for color normals to identify and correct.  

On the topic of color vision tests, an enthusiastic recommendation. Colorblind blogger David Flück provides links to three online colorblindness tests — a lightness controlled hue matching test, an online simulation of the Nagel anomaloscope color matching test, and an online simulation of the Farnsworth hue sequence test.  

Dichromat Color Appearance Simulations. A final answer to the question, "what do color deficient persons see?" is what I call the color appearance simulation, which can be caricatured as an image containing only gradations of yellow and blue color. These are in common circulation — both in everyday references and in the academic literature — primarily as visual aids to guide trichromats in designing displays or media that are equally legible to trichromats and dichromats (see for example the color simulators linked below).

A completely different usage is the reduced color image that purports to show "what colors look like to a color deficient viewer". These typically choose a festively colored image, such as the examples below.
 

red/green dichromacy color appearance simulations
(top) a normal trichromatic image, (middle) simulated deuteranomalous image (M cone similar to normal L cone), (bottom) deuteranopic image (missing M cone); images transformed using Sim Daltonism 1.0.3

 
These images can be interpreted from three very different points of view. Most simplistically, they present to trichromatic eyes the same types of color confusions that appear to dichromats. Thus, it is very difficult to accurately count the number of orange California poppies in the dichromat images, because these blend into the background textures; most dichromats will have the same difficulty. These simulations are all based on mathematical models of color vision, which are based on the experimentally observed color matching confusions of dichromats.

I believe the two other interpretive points of view are indefensible. One is that the image simulates the color sensations experienced by dichromats. This is false in the examples above, simply because most trichromats would describe the foliage in the bottom image as green yellow, when "green" is a color experience that is apparently entirely lacking in red/green dichromacies. Shifting the hue of the yellow does not resolve the objection; it merely makes it more difficult to illustrate. (Trichromats would define a pure yellow as a yellow containing neither red nor green, which are both color categories unavailable to dichromats.) Sensations, even "innate" sensation categories such as pain, are the outcome of maturation and experience, so there are philosophical problems with this position.

The other indefensible interpretation is that the images simulate the color experience or color world of dichromats. Here the objection has more to do with what might be called a feeling of color community. Trichromats share a very large community of color experience which anchors their color discriminations and color responses in all situations. Dichromats live as a minority community and must cheerfully accept the guidance and expectations of the normal majority. They must learn the objects, skills and social situations in which their color deficiency may present a hazard or awkwardness, and learn how to manage those challenges on a continuous basis. It is impossible to argue that changing the colors of a trichromatic image come anywhere close to simulating the complexities of that color world.

Two similar and very useful color experience simulators are available at the Colorblind Web Page filter and at Vischeck. Both transform the text and images of any web page into colors simulating the page experience for viewers with any of several different kinds of color deficiency.  

Yes ... but I still see all colors! Pole's description uses only two hue labels — yellow and blue. But in my reader correspondence, several color deficient artists have claimed that they can see some colors that are supposedly not visible to their type of color deficiency.

In fact, since the 1960's there have been several studies showing that dichromats offer color descriptions that are very similar to those of trichromats, especially for wide field color stimuli (larger than 4° visual angle). A recent study in the hue scaling of monochromatic (strongly saturated) lights clarifies both the perceptual and philosophical aspects of this issue.

predicted color labeling in three types of observers
proportions of four unique hues reported in the appearance of 2° wide (foveal) monochromatic lights by protanopes, deuteranopes and trichromats, at two luminance levels (230 trolands [dark lines] and 920 trolands [light lines]); from Wachtler, Dohrmann & Hertel (2004)

 
The three categories of observers, identified by genetic screening, were asked to judge the proportion of unique hues in a series of monochromatic lights. (They were free to use other color labels if they desired, but none did.) Importantly, the monochromatic series was presented at two luminance levels, to assess the effects of luminance contrast on color perception: either a 230 td light seen against a 10 td background (high luminance contrast, low average luminance), or a 920 td light seen against a 920 td background (low luminance contrast, high luminance).

First, the issue of perception. The trichromat hue labeling replicates the basic pattern reported by Hurvich & Jameson. The effect of higher luminance to increase the proportion of yellow and blue, and decrease proportion of red and green, for all viewers mimics the Bezold-Brücke effect.

Comparison of trichromat with dichromat hue perception shows several important differences: (1) for protanopes, the lack of any "red" sensation in short wavelengths (<470 nm), (2) for deuteranopes, an absence of hue discrimination in the middle to long wavelengths (>530 nm); in both protanopes and deuteranopes (3) a shift of the "yellow" transition very far into the "green" wavelengths (to around 520–530 nm) with a corresponding reduction in the range of "green" perception; and (4) a strong luminance/contrast effect on hue perceptions above ~530 nm, causing the brighter, uncontrasted color displays to appear substantially less "red" and more "yellow". The comparison also shows that (5) hues in the range 470 to 520 nm — the "X" pattern of blue and green labels — is relatively constant.

These results provide a striking illustration that hue labeling is not qualitatively linked to the L, M and S cones. A "red" perception appears in the short wavelengths (<470 nm) and across all the long wavelengths (>525 nm) in deuteranopes who are missing the L or "red" cone. And a "green" hue is perceived by dichromats only across the span of the S or "blue" cone.

What then do the dichromats identify with "green," "yellow" and "red" color labels? For reasons due to material chemistry, many substances that reflect mostly long wavelength light can show a very abrupt shift in the transition from low to high reflectance, producing what I call the "warm cliff" reflectance profile. This "cliff" reflectance characterizes all strongly saturated colors from green yellow to deep red. The cliff is generally so abrupt that it links each hue defined by a "cliff" profile to a specific lightness or contrast with white reflectance.

Trichromats perceive a yellow component in long wavelength hues at high chroma, both in lights and in surfaces. But in surface colors, the yellow component of long wavelength hues only occurs when their lightness is close to the maximum, "warm cliff" lightness possible for that hue. If the lightness is reduced, the yellow component changes to green or brown as the material color is darkened. This is most obvious to trichromats in yellows, which appear to become green when they are dulled or darkened with black pigment, and in oranges, which appear to become brown (the unsaturated color zones). Thus hues may become darker but retain their color appearance, as they do in shadows where white would appear darkened as well; or hues may become darker and lose their color appearance, as happens when they are mixed with black pigment.

In dichromats, this separation of darkening and blackening is probably disrupted, and the different confusion hues shown for protanopes and deuteranopes in the Farnsworth Munsell test is probably dependent on the lightness at which different hues are compared. In the study I have been discussing, the high luminance/low contrast condition increases the use of the "yellow" in dichromats, but luminance and contrast were not changed separately, so we cannot say how each one is related to hue naming. The crucial labeling changes would appear when the color area is darker than the achromatic background.

A phenomenological analogy for these effects is that both protanopes and deuteranopes see the world under a vibrant, monochromatic orange illuminant, and that deuteranopes also see "warm" (orange to green yellow) hues at a very similar lightness. In this world "yellow" refers to all warm colors, and "blue" is whatever distinguishes colors that are not yellow from black; "red" corresponds roughly to what trichromats would call brown, and "green" to what trichromats would call "whitish yellow"; as the tint in the illumination increases, both dichromats experience a categorical shift in the hue of green to red surfaces. Protanopes have better discrimination among yellow and red hues but this discrimination suffers from luminance changes in the color display; deuteranopes have an overall worse discrimination among yellow and red surface hues but will be less sensitive to luminance changes. (Note that saturation scaling was not requested in the study I am discussing.)

These color labeling results suggest that the color experience of trichromats and dichromats differs at a sensory level in much the same way that the color perceptions of an individual trichromat differ between a "white" and an "orange" illuminant, while the color labeling of trichromats and dichromats is similar at a cognitive level in the same way that a trichromat will comfortably label color samples as "yellow", "blue", "red", "green" or "white" even when they are illuminated by a strongly tinted orange light.

Dichromats learn a form of labeling constancy that occurs between the sensory coding of color and its interpretation in the environmental context, not between the receptor outputs and the representation of colors in consciousness. It is the link between the interpretation of the environmental context and our language structure that allows dichromats to produce trichromatically sensible color labels for color sensations that dichromats experience under a glowing gold light.

Philosophical Quibbles. The philosophical issues (as opposed to the perceptual and labeling issues) have to do with "what color is", and a common argument is that color categories are sensory code primitives, in the sense that they are not disrupted by genetic deficiencies in the receptor organ, any more than they disappear in trichromacy under strongly tinted illumination.

A surprisingly large literature has sprung up to address the question, what colors do color deficient people see? The arguments here are subtle and complex, but a few observations are worth making.

The most basic is that philosophy occurs through language, and language is fundamentally a consensual activity based on shared experience. Where shared experience is lacking, language very consistently breaks down. As each of us has no experience of the color world of any other person, we cannot reach with language to the ultimate nature of sensation. We can only observe that people agree or disagree in the labeling and discrimination of colors.

The biological basis for color vision is genetic, in the same way that (for example) the experience of pain is genetic: nearly all of us can experience pain, and on that basis it is plausible to argue that the sensory pain we experience is similar in kind, though not in quality or variety necessarily, to the pain that other people and even other vertebrates experience.

In color, the genetic contribution is strongest, and least ambiguous, in the coding of photopigments and the distribution of cone types in the retina. It is surely also very strong in the coding of the brain's neural architecture — the opponent color channels, the functional specialization of the visual cortex, the pathways between the visual cortex and other functional areas — so color must have a similar visual significance in terms of sensation and behavior. It is only when we reach the level of language and consciousness, which is a point of dynamic interdependence between color cognition and another very different form of cognition, that it gets hard to define what we can mean by "the same color" experienced in different minds.

Cognitive color and language labels are sorted out, in relation to perceived illuminance and objects in context, at the transition from sensory to cognitive interpretation. In the same way that normals learn to see and discriminate between green yellow, middle yellow, orange yellow, dull yellow (ochre), etc., but do not much mind calling all these different sensations "yellow", color deficient persons may learn to apply different color names to colors that appear to them trivially different.

Confusion lines, and the "aperture" color presentation of hues in the study described above, accurately show what color deficients can or cannot see. Color naming involves relative judgments oriented around specific anchors, for example that "yellow" is different from "blue" on a hue circle. There is little doubt that dichromats use the "yellow" and "blue" labels consistently; the labels "red" and "green" are used less consistently, with "red" in particular standing as a synonym for "dark" (as in Pole's description of the hue circle, quoted above). We can infer then that green is simply the dichromat term for bright or light valued.

Summary. The considerations I've discussed should serve as reminders of how little we actually know about color deficient vision:

(1) Colorblindness is best thought of as a difficulty with color discrimination, not an absence of color. The problem is that it is philosophically and pragmatically impossible to show that two people "see the same color" given that color is inherently a subjective sensation. But is is very easy to show that color deficients say two colors appear the same to them that to color normals appear different, or to show that color deficients cannot consistently match a certain wavelength of light with a single, specific mixture of two or three "primary" colors. And we have no useful way to describe how these confusions may affect their overall color experience.

(2) The trichromatic theory is the standard framework used to describe color deficiencies. It is essentially a retinal theory of color vision. Yet significant aspects of color perception — for example, the three colormaking attributes — only arise much later in the visual system and in the context of a "real world" interpretation of what we see. Color at this stage apparently depends to a considerable degree on the visual context and past experience in a way that allows color deficients to compensate pretty well for their color confusions. The confusions appear so clearly in laboratory hue discrimination tests because the tests are unfamiliar and the color discriminations are presented as isolated displays. (And don't forget, the discriminations are so subtle that many "normals" get some of them wrong, too.)

(3) The assumption that the retinal photoreceptors create the color sensation is physiologically incorrect; sensory qualities arise from the brain's interpretation. In the research record there are instances of colorblind persons reported to see long wavelength light as "yellow" despite the lack of a red or green cone; and the rare case of a man, a dichromat in only one eye, who claimed to see long wavelength light as "orange".

(4) It is difficult to overstate the inadequacy of language to specify colors, or the insensitivity of the average person to color differences. If the color deficient person is an artist, then discussions of color "errors" are often between people who differ in color vision capabilities, precision in the use of color words, and awareness of color nuances based on training and careful looking. It's hard to know what can be reliably defined in a conversation between these two people!

Color Deficiencies and Painting. The final question I receive is: how should an artist cope with color deficiencies?

Most of my color deficient readers recommend the use of a limited palette, in particular a selection of paints that creates minimally confusing color mixtures. It's worth considering the fact that traditional easel painters, because of their historically limited palettes, rendered colors with gamut limitations that are easily as extreme as many types of color deficiency.

However, 20th century painting styles have created and provided justification for extremely individualistic and even arbitrary color schemes in painting. The expressive and abstract uses of color mean that viewers are no longer surprised to see a blue face or a green sky. Among the paintings sent to me by persons identifying themselves as color deficients, I have not seen any uses of color that strike me as bizarre or unacceptable.

For color deficients as for "normal" artists, the same good advice applies: follow your muse, trust your instincts, and paint with feeling. Then let your art speak for itself.