294 



DR. L. N. G. FILON ON THE 



The course of the diagrams in fig. 8 is very similar. There is a well marked peak 

 about 4700, followed by a depression in the neighbourhood of 5000. There are also 

 slight indications of a depression at 6000, with a subsequent rise. 



4200 4500 5000 



WAVE-LCNCTHS IN rf/vm-Affrsrfs. 



5500 



1st order obe. 



-- Ilnd 

 Illrd 



6000 



% Tension obs. 

 Pressure 



6500 6700 



Fig. 7. Glasses 3296. Diagram of residuals. 



These systematic residuals, which are in most cases quite large, and which are 

 shown in the same place by all the tension and pressure observations of these glasses, 

 cannot be chance effects. Neither can they be affected, denoting, as they do, compara- 

 tively rapid changes in \, by any of the slowly varying corrections which have been 

 discussed. They can be accounted for only in the following ways : 



(1) Possible erroneous identification of a spectrum reference line in the neighbour- 



hood and consequent wrong determination of \ observed ; 



(2) Bad division errors of the spectroscope circle ; 



(3) Bad error in some of the weights employed at these points ; 



(4) Systematic change of personality of observer in this neighbourhood, due to 



change of colour ; 



(5) Actual variation of the law of stress-optical effect in this neighbourhood. 



(1) is ruled out by the fact that in the glass 3453, where the effect was first 



