colors in a more nearly permanent medium. Glass is a frequent choice because of its generally supeprior permanence. Some degree of metamerism has then to be tolerated because the standards have I coloring constituents not a perfect spectral match for the unknown. It is also rare that a perfect job of color matching for any standard illuminant and observer is done. The observer is then faced with what is often a difficult, and sometimes an impossible, task. He must estimate the position of the unknown color on the scale, and often it will seem to him that the unknown color is not equal to any of the standard colors, nor intermediate between any two of them. The concepts in terms of which the observer perceives these color differences then come into play. If he judges the color difference between the two luminous areas presented to him in terms of hue, brightness, and saturation, as is fairly common, he could estimate the position of the unknown color on the color scale as the point on the scale yielding the same hue, or as that yielding the same brightness or the same saturation; or he could try to estimate the point on the scale yielding the closest color match; or he could disregard brightness differences and try to estimate the point on the scale yielding the closest chromaticity match. The determination becomes an estimate based on what criterion of equivalence is used by the observer, and it depends upon his mental capability in an essentially indescribable way. In spite of these drawbacks, a good color scale is a useful time-saver, as long as it is not used in attempts to provide a one-dimensional solution to what is essentially a multidimensional problem.
Judgments of position on the color scale according to equality of brightness can be expected to correspond to luminous transmittance. Judgments according to equality of hue agree well with the Munsell renotation hue; loci of constant hue for ordinary conditions of observation are indicated on the () diagram by the curved lines of figure 3 separating the areas corresponding to the various hue luimes. Judgments according to equality of saturation agree well with Munsell renotation chroma (see fig. 22). If there is only secondary brightness variation along the scale, judgments of nearest chromaticity match may be found approximately by taking the shortest distance on a uniform-chromaticity-scale diagram [14, 23, 55, 66, 141, 142]. Figure 8 shows the uniform-chromaticity-scale triangle according to Judd [66]. If the chromaticity coordinates of a color are () in the standard CIE system, the color would have chromaticity coordinates (r,g) in this uniform-chromaticity-scale triangle in accord with eq (5). On this diagram the ellipses of figure 7 would be equal 3 tangent circles. If there is primary variation of J both luminance and chromaticity, no reliable way of estimating the nearest color match has yet been developed. According to the OSA Colorimetry Committee [24] , "The complete experimental clarification of this problem is one of the major programs yet to be undertaken in the field of colorimetric research."
Perhaps the most widely used one-dimensional color scale is that of color temperature for classifying light sources. The color temperature of a light source is the temperature at which the walls of a furnace must be maintained so that light from a small hole in it will yield the chromaticity of the source to be specified. The color scale thus consists of the series of lights producible by closed-cavity radiation and is specified by temperature on the absolute scale (degrees Kelvin). Working standards of color temperature may consist of an incandescent lamp operating at a fixed voltage combined with a series of amber or blue glasses, like the Lovibond blue glasses; but by far the most common way of producing these chromaticities over moderate ranges of color temperature is by variation of the voltage applied to an incandescent lamp. The locus of these chromaticities (the so-called Planckian locus) is shown on figures 7 and 8. If the chromaticity of the light source is close to, but not exactly equal to, any of the Planckian chromaticities, still it is possible to correlate a color temperature with the source by taking the nearest chromaticity match. Figure 25 shows this correlation [67, 81]. The isotemperature lines, which cut the Planckian locus at varying angles, are all such as to be perpendicular to this curve on figure 26. The and chromaticity coordinates of a number of Planckian radiators ( = 14,380) and standard sources in the 2° and 10° observer systems have been computed by Nimeroff [123]. These are listed in table 12. Since color temperature specifies only the chromaticity of a light, there are many spectral compositions corresponding to the same color temperature. Color temperature of a source is therefore an incomplete and unreliable indication of the rendering of the colors of objects illuminated by it or of the photographic effect of the source. To make color temperature a meaningful and useful basis for comparing two lights it must also be known that they are spectrally similar. Thus, incandescent lamps may be usefully intercompared by means of color temperature, and fluorescent lamps with about the same admixture of mercury spectrum may also be so intercompared, but incandescent lamps may not be intercompared with fluorescent.
When the degree of refinement and quality of such products as oils, rosin, and sugars may be characterized on similar one-dimensional color scales which range from dark red through yellow to perfectly colorless, the development of a complex color specification for these products is redundant. These perceived color changes correlate
35
