Color Categorization and Naming in Inherited Color Vision Deficiencies
Color categorization and naming behaviors of human observers that experience forms of color vision deficiency called “dichromacy.” Such deficiencies are sex linked and predominantly affect males and are due to errors in photopigment expression or functioning, or to the failure to inherit the genetic precursors for the expression of a normal set of retinal photopigments.
Color Naming and Categorization in Color Deficients
Color Vision Deficiencies
Only inherited forms of color vision deficiency associated with changes in the genes on the X-chromosomes will be considered here. Normal color vision is trichromatic and results from light absorption by three different types of photopigments located in the retinal cell receptors called “cones.” Short-wavelength (SW), medium-wavelength (MW), and long-wavelength (LW) cone photopigments absorb light maximally in short (440 nm), medium (540 nm), and long (560 nm) wavelengths. A genetic polymorphism affecting X-linked genes encoding for MW and LW cone photopigments induces small variations in their absorption peak, leading to subtle variations among normal trichromats; larger variations produce deficiencies. In the extreme case, a gene deletion occurs, and as a result, color vision will lose one dimension and will be dichromatic. A more moderate condition arises from gene alterations that generate hybrid photopigments. In this case, color vision is referred as anomalous trichromatic. When LW photopigment is affected, deficiencies are of the protan type, with about 1 % of dichromats (of “protanope” type) and 1 % of anomalous trichromats or “protanomalous.” When the MW photopigment is concerned, deficiencies are of deutan type and account for frequencies of 1.4 % “deuteranope” and 4.6 % of “deuteranomaly.”
In practice, color deficients confuse colors that are easily discriminable to normal trichromats. To gain intuition on color deficient’s perceptual experience, two clarifications are useful to bear in mind. The first consists in defining the term “color” which can be polysemous. Color can be defined as corresponding to the appearance of objects and lights resulting from the combination of three perceptual attributes associated to three physical variables, that is, “hue” (associated to dominant or complementary wavelength), saturation (associated to purity), and brightness (associated to luminance). Color deficients primarily confuse hues and retain the ability to discriminate between different levels of saturation and brightness. Secondly, photoreceptor absorption process is not useful when considering the perceptual implications of their dysfunction. This is better accounted for by considering further stages of color processing which involve synergic and antagonist neural linkages between photoreceptors to form an achromatic (white vs. black) [MW+LW] and two-chromatic channels: “red” vs. “green” [LW-MW] and “yellow” vs. “blue” [(LW+MW) – SW]. Dichromats who are missing either LW or MW photoreceptors are assumed to retain a single color-opponent channel: [MW-SW] or [LW-SW] (protanope or deuteranope, respectively). Anomalous trichromats possess a residual yet functional “red-green” [LW-MW] channel. Considering the diversity of anomalous trichromat types, illustrations of color deficiencies will be limited to dichromatic vision.
Despite substantial differences in color experience, the condition of color deficiency had not been reported until the end of the eighteenth century. John Dalton (posthumously diagnosed in 1995 as a deuteranope by means of molecular genetic techniques ) provided the most comprehensive description of color deficiency based on his own experience. It is at age of 28, when noticing that a geranium changed its color from sky blue in daylight to yellowish in candlelight, that Dalton suspected his color vision to be different from that of others. From systematic observations of the solar spectrum, Dalton established that if people distinguished six color categories, namely, red, orange, yellow, green, blue, and purple (further divided in blue and indigo to fit with Newton’s nomenclature), he was able to see only two or at most three distinctions: yellow and blue or yellow, blue, and purple. This description is in fair agreement with today simulations presented in Fig. 2.
Dichromats Color Naming and Categorization Abilities
In everyday life, colors are usually object colors, and as young children, dichromats can learn to associate color names with objects of predetermined color such as green grass or red cherries, and it is not surprising that dichromats use “red” and “green” to refer to colors that seem so distinctive to others. However, “red” and “green” are also used accurately most of the time for non-predetermined object colors such as clothes or furnishings, and asking color names to color deficients gives little indication of a color perception deficiency. It is in laboratory situation that one can measure the discrepancy between perception and naming in dichromats. Jameson and Hurvich  in an article entitled “Dichromatic color language: “reds” and “greens” don’t look alike but their colors do”  reported an experiment using arrangement color tests (i.e., Farnsworth D-15, with 15 colors), where colors printed on small caps should be ordered by perceived similarity. In this type of tests, dichromats make systematic and identifiable color alternations from which color diagnosis is based. For instance, dichromats put cap #2, considered by normal trichromats as “blue-green,” next to cap #13, a “violet-blue” cap. When subsequently, dichromatic observers were asked to name the same stimuli, the authors noted several interesting observations. Firstly, and despite confusing their colors, all dichromats used “red” and “green” words. Secondly, a dichromat could describe a color cap as “red-green.” This designation never occurs in normal trichromat observer as, due to their antagonist nature, these percepts are mutually exclusive. Thirdly, a large interindividual variation was reported among dichromats. For the less keen observers, “red”-“green” were used indifferently to refer to colors seen as red or green by normal trichromats. Yet, some dichromats were able to produce a naming pattern indistinguishable from that of a normal trichromat, thus revealing a gap between perceived similarity and naming. For instance, if cap #2 was placed next to cap #13 in the arrangement test, in agreement with normal trichromat, cap #2 was named “blue-light” and cap #13 “reddish blue.” In this experiment, correct naming was explained by subtle luminance differences between red and green caps that could have been used by dichromats. To the careful dichromatic observer the simple rule “darker, then red” provides a better-than-chance correct naming.
This rule would be more difficult to apply when colors are shown in isolation. This setup was used in an experiment with 140 Munsell color samples varying in hue, saturation, and brightness intended to probe naming and categorization. In the naming task, participants had to choose one of the eight chromatic basic color terms (BCT, namely, red, green, blue, yellow, orange, pink, purple, and brown) to describe each sample presented in isolation. Categories thus obtained for the two dichromats tested were very similar to normal trichromatic prototypical naming categories, with 66 % and 72 % of the samples described with the same BCT as used by normal trichromats. In the categorization task, color samples were sorted in eight categories based on their perceptual similarity with, this time, the overall sample collection displayed on a large table, inciting comparison strategy. Despite the possibility of samples’ visual comparison, local category inversions corresponding to cases where stimuli were assigned to groups in a way that is contrary to the structure of the nearest neighbor were observed in the categorization task, while absent in the naming task. Moreover, for the same number of categories, color naming produced a more consensual categorization pattern compared to perceptual similarity. Indeed, color names that define category boundaries in terms of hue (i.e., red, green, blue, yellow, orange, and purple), saturation (i.e., pink), and brightness (i.e., brown) further constrain the elicited categories by, for instance, excluding categories such as turquoise or pastel that would be legitimate based on similarity criteria. This added constraint in normal trichromats increased the consensus between participants from 72 % (categorization task) to 82 % (naming task) and similarly from 66 % to 76 % in color deficients . Language appears to operate normalization effect on categories obtained by naming in color deficients, yet it does not seem to be readily available to fine-tune color categories when these are based on perceptual similarities.
Explanations for Dichromat Color Naming and Categorization Abilities
Aside from subtle luminance differences or lightness cues, dichromat color naming and categorization performances have been explained by the contribution of other visual signals. In particular, rods distributed in peripheral retina and absent from the fovea are mediating low-light intensity vision and can act as a third photoreceptor. Under moderate light conditions, for surfaces larger than 4° of visual angle or presented in periphery, dichromats reveal trichromatic color matches comparable to normal trichromats’ small-field (2°) color match, that is, for these larger field sizes, they need two primary lights (red and green) to match a spectral yellow light. Under these conditions, dichromat large-field trichromacy is explained by rod-cone interaction providing a residual red-green sensitivity . In the naming of 424 OSA Uniform Scales samples presented in isolation, dichromats produced trichromat-like responses in conditions where rod-cone interactions were possible. When rod contribution was excluded by bleaching the photopigment by light preexposure, trichromat-like naming was no longer possible in the red-green dimension .
Dichromats still perform better than expected on perceptual tasks in conditions where luminance and rod-cone interaction are ruled out. For instance, in hue-scaling experiment, stimuli were monochromatic lights presented in a 2° field. Observers’ task consisted in describing each stimulus by a given proportion of “yellow,” “blue,” “red,” and “green.” In this experiment, a molecular genetic analysis had confirmed that dichromat participants had only one functional X-chromosome-linked photopigment opsin gene (either L or M), excluding the possible contribution of a residual red-green color vision. Dichromats reported the presence of a red component in the 420–450 nm interval and explicitly referred to blue and red mixture. These data suggest a dichromatic color space structure is richer than that illustrated by simulations based on linear model and reduction assumption as presented in Fig. 2. A model including a nonlinear transfer function of signals from S and L or M cones can account for a richer dichromatic perceptual space with enhanced naming and categorization abilities .
Non-chromatic visual cues, rod-cone interaction, or nonlinear transfer function could all provide supplementary visual information to dichromats to resolve ambiguities in an otherwise impoverish chromatic environment. However, color naming shows a more trichromat-like pattern than similarity or perceptual categorization tasks can elicit. The comparison between colors and color-naming structures has been addressed in similarity tasks using color cards and their corresponding color names printed on cards. Similarity judgments were requested for 36 pairs of stimuli. Apart from a violet-blue inversion, the multidimensional scaling analysis of dichromat color-name data provided a Newton color circle in a two-dimensional model with yellow-blue and red-green axis. For color similarities, the color circle was distorted along the red-green axis, bringing these two colors next to each other .
Dichromats have correctly learned the relational structure between color names as established by normal trichromats. Yet, as noted by Jameson and Hurvich , a dichromat’s correct conceptual representation of color is not used to optimize performance on perceptual tasks. This finding indicates that the existence of an isomorphism between percept and concept structures is not compulsory; each type of representation coexists with no apparent conflict and can be used independently to fulfill specific task demands.
- 1.Newton, I.: Opticks: or a Treatise of the Reflections, Refractions, Inflections and Colours of Light. Smith & Walford, London (1704)Google Scholar
- 4.Jameson, D., Hurvich, L.M.: Dichromatic color language: “reds” and “greens” don’t look alike but their colors do. Sens. Processes 2, 146–155 (1978)Google Scholar