May 29, 1903.] 



SCIENCE. 



867 



emitted by mercury vapor in a vacuum tube 

 made incandescent by the passage of an elec- 

 tric discharge through it, Perot and Fabry* 

 were able to secure interferences with a path- 

 difference of 43 centimeters, which is equal 

 to 790,000 wave-lengths; while Professor 

 Michelson, of Chicago, who is preeminent in 

 this work, and whose interferometer is the 

 instrument generally used in such researches, 

 informs me that he has obtained interferences, 

 with the light from this same mercury line, 

 when there was a path-difference of 840,000 

 wave-lengths. In this case the variation in 

 wave-length could hardly have been greater 

 than one part in 1,000,000 — truly an extraor- 

 dinarily close approach to perfect uniformity. 



Now, in all these experiments there was no 

 sign of the color disappearing as the wave- 

 lengths approached more and more nearly to 

 equality. Indeed, Professor Michelson's ob- 

 servation is that as the light approaches per- 

 fect homogeneity the intensity of the color 

 sensation is slightly increased! 



Light of a single wave-length in optics cor- 

 responds to sound of an absolutely pure tone 

 in acoustics. A well-made tuning-fork is by 

 no means a perfect instrument, and yet it 

 emits a note closely approximating a pure 

 tone; but such a fork is just as efficient in 

 producing the sensation of sound as the most 

 complex mixture of wave-lengths given forth 

 by any instrument. 



Is it possible, then, that these little varia- 

 tions in the wave-lengths, and not the wave- 

 lengths themselves, are the essential physical 

 cause of the sensation of color? Surely it 

 would be just as reasonable to believe that, 

 by removing all the impurities from water, 

 or nitric acid, or any other definite chemical 

 compound, these substances would lose the 

 taste characteristic of each. 



I think it unreasonable, therefore, to con- 

 tend that for the production of the sensation 

 of color it is necessary to have a superposi- 

 tion of waves of different lengths. It is quite 

 true that two color sensations which are in- 

 distinguishable from each other may be pro- 

 duced in different ways — either by light of 

 approximately uniform wave-length or by a 



* Comptes Rendus, Vol. 128, p. 1223, 1899. 



combination of quite different wave-lengths. 

 For instance, a mixture of red and green will 

 give an orange which, as a sensation, is in- 

 distinguishable from spectral orange. In 

 Maxwell's phrase, they are chromatically the 

 same, but optically they are different. In 

 such matters the eye is much inferior to the 

 ear, which can, in many instances, resolve a 

 compound sound into its constituents; and it 

 would be hard, indeed, to produce a combina- 

 tion of sounds which would so perfectly simu- 

 late a simple tone that one could not distin- 

 guish between them when they were heard 

 together. 



But I can not see why color-quality should 

 not be considered simply as a function of the 

 wave-length. Dr. Kirschmann says : " It may 

 just as well be — and the probability for this 

 supposition is even greater — that the color- 

 quality is a function of the superposition of 

 wave-lengths, so that to every qualitative dif- 

 ference in spectral colors corresponds a dif- 

 ference in the mode of superposition." I 

 think the facts I have given show conclusively 

 that a spectral color is not at all dependent 

 on any ' mode of superposition.' We need no 

 such idea to define spectral colors, and the 

 introduction of it seems a needless complica- 

 tion. 



Let us now briefly consider Dr. Kirsch- 

 mann's ' inverted spectrum,' and his applica- 

 tion of the principle of superposition to ex- 

 plain the true position of purple. 



When we view through a prism a dark line 

 of the proper width, on a white background, 

 we see a kind of 'inverted spectrum,' with 

 purple in the middle. With Dr. Kirschmann, 

 ' we must agree that the existence of this color 

 does not prove anything more than that the 

 mixture of the ends of the spectrum gives 

 purple.' That is all the experiment appears 

 to me to prove. 



If a narrow bright band be viewed through 

 a prism we get a ' pure ' spectrum. As we 

 all know, the term ' pure ' here has not a very 

 definite meaning. It signifies that certain 

 optical requirements have been complied with; 

 and if we are using sunlight the familiar test 

 for purity is the presence of the Fraunhofer 

 dark lines. If now the bright slit be gradu- 



