This is a guest post by vision scientists Qasim Zaidi, Graduate Center for Vision Research, State University of New York, and Bevil R. Conway, of the Neuroscience Program, Wellesley College. For correspondence, please email: firstname.lastname@example.org
What humans call light is the narrow band of wavelengths of the electromagnetic spectrum that we can see. Newton showed how the colors we see in rainbows correspond directly to specific wavelengths of light. Another great scientist, James Clerk Maxwell, showed that human color perception is the result of lights absorbed in three different cone photoreceptors in the retina, each sensitive to different wavelength bands. Powerful neural circuits in the eye and brain enable us to perceive millions of different colors spanning the three-dimensional space created by comparing the activity in the three cone classes. Consequently, we can distinguish more objects by their colors than most species of mammals that have two or fewer types of cones. However, with just three types of cones, for every color we can see, there are millions of physically distinct stimuli that all appear the same color. For example, yellow which is a single wavelength in the middle of the rainbow, can also be produced by combining red and green lights on a color TV.
When it was discovered that mantis shrimp have 12 or more different colored cone types, the scientific imagination ran wild. Could these shrimp see the two yellows as two different colors? Might they have a 12-dimensional color space, and be capable of seeing colors we cannot even imagine? To answer this question, Hanne H. Thoen, Martin J. How, and Justin Marshall conducted behavioral experiments with mantis shrimp to see how good their color vision really is. They report their findings in a recent article in the journal Science. Simply put, mantis shrimp are shockingly poor at discriminating colors that humans see as distinct. Thoen et al. conclude that the cone types must work independently of each other to identify colors—rather than in conjunction, as they do in humans. They conclude that mantis shrimp color vision is fundamentally different from any documented system.
The mantis shrimp results are exciting because they provide the first direct evidence of this color identification strategy, but we have reason to believe that the color-vision systems in humans and mantis shrimp may be more similar than they first appear. In trichromatic macaque monkeys, whose color-vision system is virtually identical to that of humans, the physiology and anatomy of the brain circuits responsible for color perception have been very carefully examined. This work has shown that the brain cells that are ultimately responsible for generating our experience of color are linked across several different brain regions. In one region several steps downstream of the cones, called the inferior temporal cortex, the cells are remarkably color specific. Some cells respond only to red, others to reddish blue, others to bluish red, or blue, or turquoise and so on. The specificity of the color preferences of these cells is strikingly similar to the color specificity of the different cone types in the shrimp. This analysis suggests that monkeys and humans ultimately use the same color identification strategy as mantis shrimp.
So why does the shrimp employ its peculiar color-processing strategy? The answer may be the combination of high speed and a small brain. By employing 12 color-tuned cones at the front end of the visual system, the shrimp eye provides rapid color identification because the identification process is hard-wired. The cost of early functional specialization, achieved by parallel segregated channels evident in the eye mosaic, is the requirement that the shrimp scan the scene to generate a complete representation of its visual world. The primate eye is fundamentally different from the shrimp, like a digital camera it possesses a single focusing apparatus for a dense array of photoreceptors. Using just three classes of cone photoreceptors enables the primate visual system to sufficiently distinguish the spectra of natural lights and objects, while maintaining good spatial resolution, and providing the means to identify objects by their colors despite variations in ambient lights and surrounding scenes. More classes of photoreceptors would better sample natural spectra, but would seriously compromise spatial resolution. By generating narrow-tuned color cells later in cortex, possibly through multiplicative combinations of broadly tuned cells that overlap in their color sensitivity, primate brains can still deploy similar rapid identification strategies as mantis shrimp eyes. Since inferior temporal cortex has millions of color tuned cells, they can sample the spectrum much more finely than the twelve mantis shrimp photo-pigments, therefore the same identification strategy could simultaneously provide excellent color discrimination.
Despite tremendous differences in human versus shrimp eye structure and brain circuitry, the striking similarity between the color sensitivities of primate brain color-selective neurons and shrimp photoreceptors provides evidence of a common computational strategy across extremely divergent species. This strategy may be an interesting example of independent evolutionary histories converging on the same fundamental computational principle, and may thus be worth emulating for machine vision systems designed to function in the real world.