Unit of Candle-power in White Light. G7 



The Cascade Method. 



Briefly, the course followed has been to interpose between 

 the pentane lamp and the high-efficiency sub-standards four 

 additional sets of lamps, each varying from the next in 

 efficiency (or operating temperature) by such an amount 

 that the difference in the hue of their radiations constitutes 

 a regular colour gradation in approximately even steps from 

 the pentane colour up to that of the 1*5 watts per candle 

 tungsten filament standards. These steps are marked re- 

 spectively by the horizontal lines 2, 3, 4, and 5 on fig. 1 *. 

 If, now, each such set of sub-standards is standardized 

 against the set below it — that is to say, the set whose light 

 is one stage redder than its own — it is possible to arrive at 

 a value for the final set by a series of photometric com- 

 parisons, in each of which the colour difference, although 

 perceptible, is too small seriously to trouble an observer with 

 normal sight. 



It should be said at once that this system does not eliminate 

 the personal error in colour photometry. It merely divides 

 a large colour step into a number of small ones. It is 

 known, however, that with scarcely perceptible colour dif- 

 ferences the photometric agreement between different 

 observers with normal sight is exceedingly close and the 

 measurements can be made to a high accuracy. On the 

 other hand, photometry with large colour differences, 

 although possible in commercial work, gives rise to erratic 

 aud inconsistent readings when judged from the standpoint 

 of the higher precision required for standardization. 



It is assumed in what follows that in the measurements of 

 lights of different colours the true basis of comparison is the 

 perception of the average normal eye. It is for the use of 

 the human eye that artificial illumination is chiefly required, 

 and it is suggested that two lights should be considered 

 equally intense when the illuminations produced by them 

 appear to be equal to a person with average normal sight. 



The nearest approximation which it was possible to make 

 to normal sight was to take the average of six different ob- 

 servers. Each one of these carried out the complete series 

 of comparisons, and none of them had abnormal colour 

 vision. The results of the investigation show that onlv 

 very small errors can be introduced by assuming that the 

 mean result will represent the average of a much larger 

 number of persons. 



* Set 3 will be found to differ from its neighbours by smaller amounts 

 than the other sets. It was included in the series because it corresponded 

 in colour with the ordinary carbon til anient standard lamps. 



F2 



