C. S. Peirce — Note on the Sensation of Colors. 249 



gested, make use of the principles of contrast. If any red spectral 

 light be sufficiently reduced, it will perfectly match any less 

 refrangible light. We may, therefore, say that a faint spectral 

 red m contrast with a bright light of the same kind, excites with 

 approximate purity one of the elementary sensations. The same 

 thing is true of the violet; and therefore a rich violet may be 

 taken as another primary color. In my book entitled Photo- 

 metric Researches, the printing of which is nearly complete, 

 I Bhow reason to think that the pure green has a wave length 

 intermediate between E and b. A faint green of this sort con- 

 trasted with a bright one appears as a verv bluish green, and 

 this may therefore be supposed to be the third primary color. 

 We have seen that it results from the theory that an increase 

 m the brilliancy of any light adds to the sensation nothing of 

 the peculiar color of that light, but only a certain amount of 

 the color of brightness. If this be the fact, then the photometric 

 of the eye should be the same for all colors. In 

 order to ascertain whether this is so or not, I have made a series 

 of determinations of my photometric probable error. Each de- 

 ■ii was based on twenty-eight comparisons of two parts 

 of the same colored disk. Since there were two unknown quan- 

 tities, namely, the relative brightness of the two surfaces com- 

 pared, and an instrumental constant, it follows that only twenty- 

 six observations were effective for determining the 'probable 

 error. Let R be my photometric probable error of a single 

 comparison. Then the probable error of a single determination 

 of R (which we may denote by r) would be -^gXR, or say 

 tVR. Having made a considerable number of such determina- 

 tions of R, with different colored disks, let us ascertain their 

 probable error from their discrepancies, considering them as so 

 pendent observations of the same unknown quantity, 

 and denote this probable error bv r'. If, then, R really is the 

 same for all colors, we should have 



waere m is the number of sets of experiments diminished by 1. 

 *J on the other hand, R varies with the different colors, and 

 accidentally, /'should have a larger value. The 

 lollowing are the values I obtained for R, the sum of the bright- 

 Qess of the two surfaces compared being taken as unity. 



