new experimental determinations of the chromatic visibility coefficients for their own eyes and those of one other subject. The results not only differ from Abney’s, but show such large variations among themselves as to indicate that these coefficients are subject to marked fluctuation among individuals. All three of our subjects agree perfectly on the proportions of the mixed stimuli required for the color matches (measured in energy units), but disagree on the photometry of the components. In other words, their visibility curves vary widely while their color curves remain constant. This may indicate either that the brilliance process is independent of the chromatic processes or that the latter, as constituents of the former, vary in weight without alteration in the mathematical form of their response functions. The average values for the factors obtained from our three subjects were: red, 0.370; green, 0.617; and blue, 0.012.
The stimulation values of all forms of radiant energy, simple or complex, can be completely specified in terms of the ratios of excitation of the three chromatic elementaries,—combined with one absolute measure of intensity, if the brilliance as well as the chromaticity of the color is to be taken into consideration. To arrive at the expression of the chromatic stimulation value of any given distribution of radiant energy in terms of the three elementaries, it is only necessary to multiply each of the values for the elementaries given in Table 6 by the corresponding ordinates of the distribution function throughout the spectrum and then to find the areas under the resulting three curves. The ratio of these areas determines the chromaticity of the color. The reduction of data in one system of color specification to data of another system (vide infra) can be accomplished with the greatest sureness of principle through the medium of a common expression in terms of the three elementaries. For the science of color the three elementaries are far more fundamental even than the spectrum, and the expression of the spectral colors in terms of these components (as in Table 6) is as natural as (say) the similar expression of various Planckian distributions (Table 17).
It is, of course, improbable that the curves in Fig 2 faithfully represent the actual resonance functions of the elementary chro-