|Karen K. DeValois and Michael A. Webster (2011), Scholarpedia, 6(4):3073.||doi:10.4249/scholarpedia.3073||revision #91140 [link to/cite this article]|
Light can vary in both wavelength and intensity. Color vision is the ability to make discriminations based on the wavelength composition of the light independent of its intensity. Color vision is distributed widely throughout the animal kingdom, and it appears to have evolved independently multiple times. This suggests that it serves important purposes. Color vision is used to determine the location and shapes of objects (e.g., fruit among foliage) and their identity and characteristics (e.g., what kind of fruit and whether it is ripe). It is particularly useful in cluttered natural scenes, where intensity variations may arise from either shadows or object borders.
The stimulus for color vision
Few natural objects emit light; most are visible because they reflect light. The proportion reflected varies with wavelength, defining the surface reflectance function. The light reaching the eye is a product of the surface reflectance function and the spectral power distribution of the illuminant. If color is to provide a reliable cue to the surface, the visual system must compensate for changes in the illuminant. Object colors do tend to appear unchanged under different lighting, showing that color appearance is closely tied to surface reflectance. Many factors are thought to contribute to such color constancy. One factor is that surface reflectance functions, and the spectral power distributions of illuminants, typically show broad, smooth variations with wavelength. Consequently natural spectra can often be well characterized without representing the energy at each wavelength ( Figure 1 ).
The human visual system samples the visible spectrum (roughly 400 to 700 nm) with a mosaic of three classes of photoreceptors, each sensitive to different but broadly overlapping wavelength ranges. The receptors have peak sensitivities to short (~440 nm), medium (~535 nm), or longer (~565 nm) wavelengths, and are called S, M, and L cones, respectively ( Figure 2 ). The different spectral sensitivities are determined by the specific photopigment molecule each contains. No individual receptor type can differentiate between a change in wavelength and a change in intensity. Although the probability that a given photon of light will be absorbed by the photopigment depends on its wavelength, all subsequent events within the receptor are independent of wavelength (known as the principle of univariance). Thus wavelength information can be extracted only by comparing the responses across different classes of receptors.
The theory of the trichromatic nature of human vision is usually attributed to Young and Helmholtz in the 19th century though earlier accounts exist. The first evidence for trichromacy came from studies of color mixture. Any light can be matched by a suitable mixture of just three primary lights, as demonstrated by the 19th century physicist Maxwell. The match occurs when the responses to the light and the primaries are equated for each class of cone. Such lights that are physically different but perceptually identical are known as metamers. Color matching functions have been successfully used to estimate the spectral sensitivities of the cone photopigments; more recently these have been assessed more directly through physiological measurements.
Color matching functions are also important because they allow one to predict the matches to any arbitrary light. All lights can therefore be represented by their coordinates in a three dimensional space whose axes are defined by a set of primaries. Two dimensional “chromaticity diagrams” are constructed by using the relative values of two of the primaries (e.g. dividing each primary by the sum of the three primaries). Color spaces are widely used in both industry and science for specifying colors. The primaries may be imaginary (in that they do not correspond to any real physical light) as in the CIE chromaticity diagram, or may be chosen to directly reflect the relative excitation of the cones as in the MacLeod-Boynton diagram (see Figure 4 ).
The S, M, and L cone types differ in number and in their retinal distributions ( Figure 2 ). The S cones make up only about 5% of the total. They are essentially absent in the central 1 deg (foveola) of the retina, reaching their peak density at an eccentricity of about 1-2 deg, and falling in density with increasing eccentricity. The L and M cones are more similar in peak sensitivity and are sensitive to all visible wavelengths. Their density peaks in the foveola and falls rapidly with increasing eccentricity. The S cones, however, are only minimally sensitive to longer wavelengths; thus, in the longer wavelength region color vision is essentially dichromatic. The S cones appear to function almost exclusively in color vision, providing little or no input to the encoding of luminance ( Figure 3 ).
The three cone classes also differ genetically. The gene encoding the S-cone pigment is only about 45% homologous with those coding the L and M pigments, and it is located on autosomal chromosome 7. The L and M pigment genes are at adjacent locations on the X chromosome and are 98% homologous. This is taken to suggest that our color vision evolved in two stages. Initially primate color vision was based on comparing signals from the S cones with those from a single, longer-wavelength pigment. Such a dichromatic system is characteristic of most mammals. More recently, the L and M cones differentiated, allowing a second chromatic axis of comparison. The L vs. M chromatic axis may be particularly useful for distinguishing ripe fruit from foliage and recognizing social signals among conspecifics. Analyses of the photopigment genes have also revealed common polymorphisms of the L and M pigments leading to variations in normal color vision. Although some individuals may express more than three cone photopigments – particularly female carriers of color deficiencies - it is not yet clear whether this can result in tetrachromacy.
Color blindness typically results from the loss of one cone type (dichromacy) or a change in the spectral sensitivity of one of the cone types (anomalous trichromacy). The incidence of color blindness varies with ethnicity but affects approximately 8% of the Caucasian male population. Most forms reflect inherited defects in the L and M photopigment genes, though acquired color blindness from disease or poisoning is more likely to affect S cones. In rare cases extreme color blindness can also result from central cortical damage (see Cerebral Achromatopia).
Trichromacy determines which spectral stimuli can be discriminated, but it says nothing about hue (that characteristic that is typically referred to as color—e.g., red or blue). The 19th century physiologist Hering first pointed out that there are four, not three, hues that are perceptually unique. There exists for each normal observer a red, a green, a blue, and a yellow that appear to contain no admixture of any other hue. All other hues can be acceptably described as a mixture, or combination, of two of the unique hues (e.g., orange is a combination of red and yellow). Further, the unique hues form two opponent axes, red-green and blue-yellow. Red and green, the two poles of one axis, cannot be seen at the same time in the same place; nor can blue and yellow. These are mutually exclusive. Any other combination (such as red and blue, or green and yellow) is possible.
Though trichromacy and opponency were once considered to represent competing models of color vision, it is not difficult to reconcile them. Trichromacy reflects limitations imposed by the number of photoreceptor classes, while opponency reflects how signals from the cones are compared. The three receptor types do not send separate and independent signals to the brain. Rather, their outputs are immediately compared within the retina ( Figure 3 ). Some neurons compare (i.e., subtract) the M-cone outputs from the L-cone outputs or vice versa, forming an LM opponent axis. Similarly, the output of the S cones is compared to the combined outputs of the L and M cones, to form an S-opponent axis. Such an opponent neuron will give an excitatory response to lights from a restricted spectral region, and an inhibitory response to lights from a different spectral region. Spectrally opponent neurons were first demonstrated by Svaetichin (1956) in the retinae of fish and by De Valois and colleagues (1958) in the lateral geniculate nucleus of primates.
Although the existence of two physiological opponent axes (with four poles) is functionally similar to the four unique hues and two opponent axes of Hering, the spectral sensitivities are not identical ( Figure 4 ). For example, a variation between unique blue and unique yellow does not isolate the responses of cells tuned to the S-opponent axis. This suggests that there are further transformations of color coding. Although much is known about the color selective responses of cortical cells, the neural basis of color appearance remains poorly understood.
Color appearance has also been examined by comparing color naming patterns across cultures. In 1969 Berlin and Kay reported that different languages parse the spectrum in similar ways, revealing strong universal tendencies in the spectral locations and relative order of basic color terms. However, cultural and environmental factors may also influence the categorization of color.
Spatial and temporal factors in color vision
Color perception is strongly affected by context. A gray square can appear reddish or greenish when viewed within a green or red surround respectively, a phenomenon known as chromatic contrast. For finer spatial patterns the color can instead appear more similar to the surround. As a result the same spectral stimulus can take on very different hues depending on the setting ( Figure 5 ). Color appearance can also be strongly affected by prior exposure to stimuli ( Figure 6 ). Processes of adaptation constantly adjust visual sensitivity according to the stimulus the observer is currently viewing. A red light adapts the L and M cones more than the S cones. This alters their relative responses to a subsequently viewed light, causing a stimulus that previously appeared white to appear greenish. A striking demonstration of color adaptation is provided by the Lilac Chaser illusion, in which lilac spots fade with viewing so that only the transient green afterimage when each spot is removed is visible. Adaptation effects occur throughout the visual system and at many time scales, and thus adjust color perception to many different aspects of the visual world. These contextual processes play an important role in contributing to color constancy.
Detailed reviews on many topics in color vision can be found in the following sources:
Chalupa, L.M. and J.S. Werner, eds. The Visual Neurosciences. 2003, MIT Press: Cambridge.
Gegenfurtner, K.R. and L.T. Sharpe, eds. Color Vision: From Genes to Perception. 1999, Cambridge University Press: Cambridge.
Kaiser, P. and R.M.B. Boynton, Human Color Vision. 1996, USA: Optical Society of America.
Mausfeld, R. and D. Heyer, eds. Colour Perception: Mind and the Physical World. 2004, Oxford University Press: Oxford.
Shevell, S.K., The Science of Color (2nd Edition). 2003, Oxford: Elsevier.
The following sources provide standard data sets for color vision:
Colour and vision database (http://www-cvrl.ucsd.edu)
Wyszecki, G. and W.S. Stiles, Color Science. 2nd ed. 1982, New York: John Wiley and Sons.