72 Messrs. Paterson and Duckling on the 



somewhat whiter than that from the pentane lamp. At the 

 time, however, that the determination was made their colour 

 formed a suitable intermediate step between that of the 

 pentane lamp and the usual light sources which had to be 

 tested. After eight years' use of these lamps, and in view 

 of some small unexplained differences between the N.P.L. 

 unit and that determined in other quarters, it was considered 

 desirable to realize the unit anew. Set No. 1 of the sub- 

 standards was therefore prepared, whose light was identical 

 in colour with that from the pentane lamp. These lamps 

 were compared against the pentane standard in the usual 

 way over a period of about two years (1912-14). The 

 results were analysed by the method of least squares in 

 order to determine the true value of the sub-standards when 

 the pentane lamp was burning under standard atmospheric 

 conditions. It is satisfactory to find, from the subsequent 

 inter-comparison of sets 1 and 2, that the unit of candle- 

 power realized on these two different occasions, separated by 

 a period of eight years, agreed with each other to within 

 less than 1 part in 1000 (0' 7 per cent.). The value of the 

 unit originally realized in set 2 is that which formed the 

 subject of agreement between the laboratories of this 

 country, France, and the United States in 1908. The 

 value obtained in the original determination has, therefore, 

 now been fully substanliated by the new determination, and 

 set 1 of the sub-standards becomes a fundamental set whose 

 constancy is well assured, and the determination of whose 

 candle-power values is independent of any colour difference. 

 Incidentally, it has been possible to redetermine the con- 

 stants of the pentane lamp affecting its changes with 

 varying atmospheric conditions. These results are discussed 

 in detail later. 



It must be admitted that it is to some extent a matter of 

 good fortune that the two independent determinations of the 

 unit have agreed to within such close limits, since the first 

 determination was made by one observer only, and since 

 there existed a small colour difference his readings might 

 well have differed from the average by a few parts in 1000. 

 It so happens, however, that his difference from the mean 

 was in this instance negligible, and the magnitude of the 

 unit as determined ten years ago remains unchanged. 



The following points of general interest arising from these 

 and allied measurements will now be discussed. 



1. The order of agreement between different observers in 



