Color Vision: A New Understanding

John A. Medeiros

By all accounts, the mystery of how the eye sees color is solved.

Yet, contradictions, puzzles, and enigmas about human color vision persist.

And how can there be no connection between form and function, cone structure and color perception?

Perhaps it is time to consider another approach to understanding color vision.

With the press of everyday affairs, it's quite natural to take our sense of color vision for granted. And yet, when we do stop and take the time to think about it, like we might when we see a particularly awesome sunset, it’s hard not to marvel at the richness and depth our color sense adds to the visual experience.

Philosophers, scientists, and laymen alike have all long wondered how color vision works. Many of the giants of scientific thinking, including Newton, Goethe, Young, Maxwell, Helmholtz, and many others have explored various aspects of vision and pondered the question of just how the eye sees colors and all have contributed in important ways to our understanding of the process.

Well, here we are in the Twenty-First Century - the age of supercomputers, molecular genetics, the Internet and space travel so by now we surely know the answer to just exactly how we see colors. Do we not?

What About Our Understanding of Color Vision?

The standard, near universally accepted, Young-Helmholtz trichromatic model explains color vision through the identification of three cone types which explains at a stroke the fundamental fact of trichromacy. Supporting this model, opsins (the protein portion of the photosensitive pigments) that are sensitive to different parts of the visible spectrum have been genetically isolated.

But there are a number of aspects of color vision about which the standard model says nothing, or indeed, says the wrong thing. Consider just ten such items:

The identified photopigments in the human eye do not correspond very well to the supposed primary colors, nor are there exactly three of them.
Co-expression –- Multiple photopigments have been identified within the individual cones of many species with no apparent problem for color discrimination function. There have even been reports of such co-expression in the cones of the human retina.
Cones are universally conical in shape. There is an absolute dichotomy in shape between the rod photoreceptors that provide black and white night vision and the cone photoreceptors that provide color vision in day light. This dichotomy in structure has long been a total mystery and has never been tied to any functional difference in the two receptor types. In addition, true monochromats (black/white colorblind) have been found, on autopsy, to universally have abnormally shaped cones.
Well-ordered color sensations (subjective colors) are induced by time-modulated purely black and white patterns (e.g., Benham’s Top). The correspondence between the color induced and the time coding is universally the same for all color normal observers.
Violet -- the shortest visible wavelength -- looks like purple; go past blue on the spectrum and at a certain point it's as if you have suddenly added a red color from the opposite end. The near identical appearance of violet and purple has long been a profound puzzle in terms of any model of color vision yet proposed. Attempts to explain the appearance of the red component at short wavelengths have all presumed the existence of a subsidiary, short wavelength absorption maximum of the "red" pigment sequestered in the "red" cones. In fact, no such secondary short-wavelength maximum has been demonstrated for any known photopigment of vision.
Two varieties of color-blindness -- supposedly the absence of red or green (L and M) cones-- are quite common, but what would correspond to absence of the third type of cone (tritanopia) is incredibly rare. There has long been controversy over whether the common red-green deficit vision is a result of missing red or green cones or else due to the wrong pigment being substituted in a given cone type. Experimental evidence exists contradicting both explanations.
Color-blind subjects can (for brighter or larger patches of color) distinguish and correctly name red and green colors, even in cases where they demonstrably lack the genetic machinery for one of the red or green photopigments.
As predicted by the standard model, a pure yellow, say, can be matched by a mixture of red and green. Startlingly, the pure yellow and the matched mixture can be distinguished dynamically: a bar of mixed red and green light moved across the retina resolves into red and green leading and trailing edges,while a moving true yellow bar does not.
There is a linear relationship between the wavelength of light and its time of perception: red is seen faster than green, which is in turn seen faster than blue.
Eye movements are necessary for vision. When vision is fixed (stabilized), visual perception, especially color vision, is promptly lost.

All these facts are mysteries -- or even paradoxes -- under the Young-Helmholtz model. However, they are unproblematic (in some cases, required) for an alternative model of how color vision works, what is here called the Cone Spectrometer Model (CSM).

Consider the fact that the cones are just the right size to serve as effective waveguides (optical fibers) to carry light from any part of the visible spectrum. Optical waveguides transmit light along their length in discrete waveguide modes which essentially correspond to light propagating along the fiber by bouncing at specific angles of reflection with the fiber wall and surround interface. Now, for appropriately "sized" cones, due to the tapering which gives cones their name, successively deeper (narrower) parts of the cone can't carry light of longer wavelengths. Red light only fits into the wide end of the cones, and travels but a short distance before being "squeezed" out by mode cut off; green light gets deeper into the cone, and blue light can shine all the way to the bottom.

The accompanying photograph shows a highly-magnified view of this effect occurring in a small tapered glass fiber immersed in a liquid with a refractive index only slightly smaller than that of the fiber. Long wavelength light is seen to leak out first and progressively shorter wavelengths are excluded at successively smaller portions of the cone. Actually, it is evident that this happens more than once in the fiber. Near the top where the initially white light is incident, the colors leaking out and visible along both "edges" of the first third of the glass fiber are first white, then rose-colored, then greenish to blue where the second-order mode is cutting off first. There is then a well-order red through blue dispersion of the spectrum along the last two-thirds of the photograph as the lowest-order fundamental waveguide mode cuts off along the smallest part of the taper.

Now light, of course, shines into the cones at the speed of light, but once a detection event happens, the resulting information travels much more slowly as a nerve impulse. The cones are, somewhat perversely, "wired backwards", so that the red-detecting, wide ends are effectively closer to the cone output at their synapse with the bipolar cells. Detection events at the deeper, blue end must propagate back up the length of the cone at this slow speed; as a result, detection of red light is signaled earlier than blue light (and in general, detection time will be proportional to how deep the detection happened). In other words, the taper cutting off different wavelengths at different depths corresponds to differences in timing (of light detection information reaching the brain).

Effectively, the shape of cones sorts wavelength information into a difference in timing; I propose that this is the essential mechanism of color vision.

In this context, what does the CSM model say about the ten problematical items mentioned above?

Identified photopigments. The proposed color detection mechanism is inherently indifferent to the details of any photopigments in the cones. Light only needs to be absorbed at a specific location along the cone and the photopigment(s) simply initiate the transduction of light into an electrical signal.
Co-expression: Multiple photopigments could actually enhance the operation of the proposed mechanism. While not necessary for its function, it would be more efficient to have long-wavelength absorbing photopigment in the broad entrance end of the cone and shorter wavelength absorbing pigment in the narrower, distal end of the cone.
Cones are universally conical in shape: The cone shape itself is the element providing the spread of colors along its length. Structural differences are thus intimately tied to functional differences between cones and rods. Absent the conical shape, the "cones" would not then be able to provide color discrimination -- just as observed in true monochromats with abnormally shaped cones.
Subjective colors: The pattern of achromatic illumination in Benham’s Top and the so-called pattern-induced flicker colors (PIFCs) exactly mimics the temporal order of the proposed mechanism.
Similarity of violet and purple: This apparent “closure of the color circle” is directly predicted by the CSM model as a consequence of second-order waveguide mode propagation. Sufficiently short wavelengths of light (violet) can excite both of the two lowest-order modes. The portion of violet light propagating in the higher-order mode will cut off like red light while the portion in the lowest-order mode cuts off more slowly like blue light so that violet will behave like a purple mixture of red and blue light.
Two common varieties of red-green color-blindness (protanopia and deuteranopia): the CSM model directly accounts for these in terms of mistuning of the basic mechanism whereby cones that are too small will be “red-blind” corresponding to the protanopic version of color deficit vision and cones that are too large will correspond to the “green-blind” deuteranopic version.
Residual color discrimination in “color-blind” subjects: lack of a red or green absorbing photopigment would not disable color function in the CSM model, although mistuned cones would certainly be less effective in color discrimination.
Dynamic breakdown of statically established color matches: While this is a critical issue with strong implications for the validity of the traditional Young-Helmholtz model (see below) it is directly in accord with the time-ordered color information of the CSM model.
Linear relationship between wavelength and its time of perception: This too is a direct consequence of the basic CSM mechanism.
Eye movements are necessary for vision: The microsaccadic eye movements are of just the right amplitude and frequency to provide the necessary synchronization signal to read the time-ordered color information established by the cone shape. On each movement of a color border across a cone, the change in cone output will be temporally correlated with the change in color across the border.

There is actually a good deal more the proposed model says about these and many other aspects of color vision. The following sections explore many of the related issues more systematically to build what is arguably a compelling case for the CSM explanation of color vision. The goal is nothing less than the dispelling of what has been many of the long persistent contradictions, puzzles and enigmas about human color vision.

1. The Standard Three-Cone Model of Human Color Vision

2. The Three-Cone Model is Experimentally Falsifiable

3. Multiple Pigments

4. What Does a Color Vision Model Have to Explain?

5. What About the Cone Shape?

6. The Cone Spectrometer Model (CSM)

7. Hue Discrimination and the Similarity of Violet and Purple

8. Color Deficit Vision

9. Summary, Final Note

References

About the Author

There exists a vast body of scientific research and published literature on human color vision. In part, this is a reflection of the natural interest in the functioning of one of our most profound and treasured human senses. It is also a reflection of the confounding complexity of the perception (involving psychology, physics, neurophysiology, biochemistry, and psychophysics) and the lack to date of a truly comprehensive model to explain the array and diversity of the myriad aspects of color vision. Given the vast scope of phenomenology involved, it is not possible to address all of it here. I do cover more details in the book, Cone Shape and Color Vision: The Unification of Structure and Perception (click for table of contents) although there is still much more to explore. The book is available as a black and white (!) paperback or as a downloadable, color PDF file from all the major online booksellers. A link to the book at Amazon.com is here.

Incidently, I have all the content of the document comprising this websie available as a single, downloadable PDF file here (about 3.4 MB).

Further information can be found on a Comments Page that I'll try to maintain with both any continuing updates and with any interesting comments (good or bad!) I might get in respomse to this website.

Email any comments to john.medeiros@mac.com.