TRISTIMULUS COLOR MEASUREMENTS IN FADING STUDIES .383 series made using a yellow aluminum lake. These compacts have had a finite change in their chromaticity although there is essentially no change in their lightness values. Table IV shows the same type of data as was previously shown in Table II for the three sets of green tablets. It may be seen that there is very little indication of a change in lightness which would be expected in a case of "fading." Table III includes the calculated values for the red and yellow tab- lets. It can be seen that when more than one color is being evaluated there is a rank inversion using absorbance units, and the K/S values fail to indicate the magnitude of the color change and its type, that is, chromaticity or lightness. Another way of illustrating this point of considering color as more than colorants is shown in Table V. It lists the calculated number of Langley units that the tablets would have been exposed to when the K/S value had dropped to 75% of its original value. The ranking has not been changed, which indicates that qualitatively this is still a satis- factory procedure. However if the Ae value, the total color change, is calculated for this period of exposure it can be seen that a 25% reduction in the Kubelka-Munk value does not result in the same degree of color change in every case. CONCLUSIONS It is suggested that fading studies be conducted as follows. The particular formulations under consideration, not the colorants as iso- lated entities, should be exposed to the desired lighting conditions. These can be either normal or exaggerated conditions. Samples should be withdrawn at periodic intervals, with the time of exposure being noted. Samples should not be allowed to fade to a washed-out color or even close to this. All small color difference formulas are invalid if the color differences are too great. Table V Relative Light Stability of the Colorants Langley Units Ae Tablet at K/S75% at K/S75% P3 •o 0.0 Yellow 620 6.2 Red 150 3.9 3D4 111 8.2 5D5 64 9.7

384 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS Spectrophotometric curves can then be run on the samples, as has been discussed here, and the chromaticity coordinates and lightness values obtained. A look at the samples, their curves, and the numbers obtained should then indicate which technique should be used to follow the course of color changes and to select the most stable colorant. It may be of value to plot color change, as zXe, versus Langley units or time, and then determine how long an exposure it takes for the sample to reach a point of five color-change units. In any case it cannot be overstressed that there must be a visual cor- relation between the ranking given by an observer and whatever sys- tem of measurements is chosen. Any criterion which disregards this is unsatisfactory. (Received November 30, 1966) REFERENCES (1) Nickerson, D. J., Seeing colours, J. Opt. Soc. Am., 51,465-466 (1961). (2) MacAdam, D. L., Visual sensitivities to color differences in daylight, Ibid., 32, 247-274 (1942). (3) Mackinney, G., and Little, A. C., Color of Foods, The Avi Publishing Co., Inc., Westport, Conn., 1962, p. 188. (4) Mudd, J. S., Precise measurement of fading on a time-intensity basis, Y. Soc. Dyers Colourists, ?3, 47-52 (1957).