0. N. Rood— Color System. 267 



about the first position for discs that have nearly the same 

 coefficient, but are quite removed from being complementary. 

 As soon as the compound disc is set in rotation it will be 

 noticed that the mixture-tint furnished by the larger disc is the 

 least colored of the two, and the amount of yellow in it is to 

 be increased, the smaller disc following with the same interval, 

 and this proceeding is to be pushed forward till the observer 

 is unable to decide whether the larger or the smaller disc is 

 the paler, the least colored. Both will differ sensibly from the 

 pure gray mixture, and the two pale mixtures will differ from 

 each other in hue, still after some practice the observer will 

 learn to decide with promptitude that they differ about equally 

 from gray. This is a much more certain and easy process 

 when in general only a little color is present, than when the 

 hues presented are at all decided, and this is the reason that 

 in building up the system the advances were made by small 

 steps. The first half of an observation being now made and 

 the mean noted, the yellow element is made from the start to 

 preponderate considerably, the interval is preserved, and with 

 a reversed approach the best neutrality obtained. The two 

 readings constitute one complete observation, and if the maxi- 

 mum reliability is demanded, not more than five complete 

 observations in any one case should be made on the same day, 

 it being more profitable to devote the remaining time to other 

 determinations. 



As an example of the results to be attained I give three 

 sets of determinations for five discs with rather low coefficients, 

 each number being the mean of only three complete observa- 

 tions made on the same day : 



CB9a -585 -575 -585 



Y6i-_. '506 -480 -486 



CB8 -466 -442 -452 



Y6£ -432 -421 '428 



B28 -.394 "384 -398 



When it is remembered that in photometric observations 

 involving light of the same color the error in any one observa- 

 tion may easily be two per cent and more it will be seen that 

 these results are satisfactory so far as uniformity goes. 



By the methods above indicated I obtained the coefficients 

 of a series of slightly differing discs ranging in color from 

 orange-yellow through red into purple, and on the other hand 

 from a somewhat yellowish green through cyan-blue to blue. 

 Then, instead of extending the process all around the circle, a 

 Newton's diagram was constructed using the standard ver- 

 milion disc, the extreme green attained and the blue. The 



