Robert W. Burnham

Comparison of Successive with Simultaneous Color Matching

The two matching methods were compared to see whether they gave substantially different results. Such a difference would be of interest because successive matching is common in everyday life whereas simultaneous matching is standard colorimetric practice. An experiment was performed in which 25 test colors were matched by both methods. The successive or memory method yielded (1) the higher variability of replicative matchings, (2) the shorter matching times, (3) systematically higher purities, and (4) somewhat higher luminances. Three supplementary experiments are cited which are confirmatory with respect to the principal finding of extra purity and luminance usually required for the memory matches. This increased apparent strength of the remembered colors seems to be a direct consequence of the selectivity of memory.

Prediction of Color Appearance with Different Adaptation Illuminations

Specifications are reported of colors which match in appearance with relatively complete adaptation to daylight, tungsten, and a third greenish illuminant. A computational procedure is described for predicting the color appearance of everyday objects. This procedure is based on the known spectral character of the given object in relation to the appearance data presented here. Certain implications for color vision theory are included.

Influence on Color Perception of Adaptation to Illumination

Six experienced observers made consistent determinations of various colors which appeared the same with adaptation to tungsten light and to artificial daylight. These observations were made with each eye viewing a different color patch and with the patches appearing juxtaposed at the middle of a fused binocular field. The method was to make the two juxtaposed patches match by adjusting one of them, sometimes when both eyes were adapted to the same illumination and sometimes to the different illuminations. Plots of the data in the CIE chromaticity diagram indicate a systematic shift in color appearance toward the blues when adaptation was changed from daylight to tungsten; or toward the yellows when adaptation was changed from tungsten to daylight. The magnitude of this color shift was substantial, at least in the considerable color region investigated, for here the average length of the representative vectors was 0.10 in CIE terms or of the order of 20 just perceptible color differences. Qualitatively, the results confirm those of Hunt and of Winch and Young. The theoretical implications will be discussed in a later paper.

Threshold and suprathreshold perceptual color differences.

Observers matched a series of colors. They also matched a given neutral difference by producing, with a colorimeter, a color difference for one color attribute at a time (producing a saturation difference while keeping hue and lightness constant, producing a hue difference while keeping lightness and saturation constant, and producing a lightness difference while keeping hue and saturation constant). We found it possible for observers trained in color scaling to abstract the individual color attributes of hue, saturation, and lightness; that a unit suprathreshold perceptual color ellipsoid can be described about a given color in color space; that the precision of color-difference matching seems to be a function of the size of the perceptual color interval of the reference colors; and, finally, that there is a tendency for color-difference-matching ellipses to orient toward the nearest colorimeter primary.

Color Constancy in Shadows

Because color samples perceived as surfaces or objects tend to look much the same under various viewing conditions, they are said to possess color constancy. The viewing condition chosen for study was an obvious shadow of daylight quality falling on a color sample and part of a surrounding white field; the less the effect of the shadow on the color appearance of the sample, the greater would be the color constancy. The purpose of the study was to make evaluations of the color constancy, both over all and by attributes, of ten color samples viewed one at a time under the standard shadow. The method was to present the color samples in the surface mode of appearance and to match them with a calorimeter, the field of which was also perceived in the surface mode. In some trials the shadow was present, in some absent, and in others there was no perceived shadow but rather the sample luminance alone was reduced proportionally. These match data were converted to the Munsell system of renotation. Brunswik-type constancy ratios were formed in terms of Munsell hue, chroma, and value taken separately. The results indicate the constancy of the hue, saturation, and lightness of the surface color perceptions. There was evidence of considerable constancy in all three attributes. After weighting the data for each attribute in accordance with an appropriate color difference formula, an estimate of the combined or over-all color constancy was obtained.

Comparison of Color Systems with Respect to Uniform Visual Spacing

The present study is a comparison of ten systems of color specification, for a restricted chromatic area around the neutral at a single luminance level, with respect to their relative adherence to uniform visual spacing. A sample of 20 visually equi-spaced (Munsell renotation) chromaticities at the same luminance were used. Chromatic coordinates for these colors were expressed in each of the ten systems and plotted in their respective chromaticity diagrams. Then all data were reduced to the same scale, after which radial saturation distances and hue angles were compared to those in the Munsell renotation criterion. Rank orders of departure from the criterion were determined for hue and saturation separately. Results show that all the 10 systems include visually significant departures from a uniform color space. These results may be generalized beyond the present sample because, if other more saturated colors were added to the present sample, there would be even greater departures from uniform spacing. The use of any particular system in any particular instance depends on circumstances, the time and personnel available, the number of points to be transformed, and the accuracy with which it is desired to illustrate visual relationships. Where the accurate expression of visual relationships is of prime importance, Munsell renotation is recommended because it was designed specifically to show equal visual intervals.

Color perception in small test fields.

Several small test colors were varied in size and shape to discover possible effects on their color appearance. Color appearance was evaluated calorimetrically with a binocular septum viewer; the test color was viewed with one eye under the given conditions and matched with a variable mixture presented to the other eye under standard conditions. Increasing the area of the test colors from 25 to 500 sq min of visual angle evoked some significant increases in saturation and characteristic shifts in hue. These changes in perceived color were found to be independent of marked changes in rectangular shape. The observed changes are related to small area tritanopia.

Predictions of Shifts in Color Appearance with a Change from Daylight to Tungsten Adaptation

Comparisons are made between shifts in color appearance reported by Wassef from W. D. Wright’s laboratory and those predicted by the Burnham, Evans, Newhall (B-E-N) equations for a change from daylight to tungsten adaptation. Prediction equations based on Wassef’s data are also reported, and a comparison is made between the B-E-N and Wassef predictions for all of the Munsell colors specified by Kelly, Gibson, and Nickerson. Both sets of data lead to similar predictions, and charts are presented from which predictions for a wide variety of colors may easily be compared with the shifts in color appearance found in practice.

A color memory test.

A test of color memory has been developed using chips from the Farnsworth-Munsell hue series. An individual is asked to select from a hue-circle of color samples the one which most resembles a test sample presented a short time before. This procedure is repeated for a number of test samples so that the observer can be reliably scored on a basis of demonstrated accuracy.

Exploratory Investigation of Perceptual Color Scaling

A viewing situation of moderate complexity was used in the present study. Two groups of color samples were used, one with constant hue and one with varied hue. Ratio judgments of color difference were obtained by having observers set the physical distance between pairs of color samples to represent the ratio of the size of visual differences relative to a standard difference created by setting two other color samples a fixed distance apart. The scaled color judgments were subjected to the Shepard–Kruskal nonmetric technique of multidimensional analysis. A comparison was made between the nonmetric analysis and more-conventional metric analyses where comparable scaling results were obtained. The outcome of this comparison was that with a nonmetric technique, and with an intrinsically imprecise scaling technique, meaningful metric visual-scaling information became available. Good agreement was found with Munsell scales and with Godlove and MacAdam predictions of visual distance.