314 



Prof. W. M. Hicks on the 



taken that the "96" represent the real relation. Taking 

 this into consideration we find the following sets of pairs : 



1, 



17604-89 (-3) 



22-84 (-9) 



40240-80(1-6) 





6755 



7361 



7966 



2 



17672-44 (-15) 



96-45 (-5) 



4032046 (-8) 



3, 



17705-26 (-15) 



21-34 (-9) 



40137-43(1-6) 





7283 



7545 



7797 



4. 



17778-09 (-15) 



96-79 (-5) 



40215-40 (-8) 



5, 



17931-65 (-16) 



21-04 Co) 



39910-44 (-8) 





7242 





-6933 



6. 



18004-07 (-16) 



22-59 (-5) 



39841-11 (-8) 





7343 



7670 



8000 



7, 



18077-50 (-16) 



99-30 (-5) 



39921-11 (-8) 



also the 



single sets 







8, 



17638-43 (-3) 



96-21 (-5) 



40353-99 (-8) 



9, 



17814-57 (-15) 



96-32 (-9) 



40178-07 (1-6) 



10, 



17842-93 (-15) 



96-63 (-9) 



40150-33 (1-6) 



11, 



17858-68 (-16) 



9S-05 (-9) 



40137-43(1-6) 



12, 



18082-43 (-16) 



96-43 (-5) 



39910-44 (-8) 



13, 



1824049 (-16) 



99-85 (-9) 



39759-22(1-58) 



Consider, first, the second pair. The separation of the 

 means is very nearly the correct value 75*20. But those of 

 the F and F doublets are not the same. This means that the 

 sequents are different in the two sets, but the same in each. 

 They are given by i(F— F), viz. 11216-08 for F] and 

 112*18'65 for F 2 . These differ by 2*57, which should be 

 explicable by displacement. Now the sequents correspond 

 to denominators 3*1245, on which the oun produces a chunge 

 •751 in the sequent, so that three ouns give 2 25. This is 

 the same as the observed within narrow observation errors. 

 There can be little doubt that this explanation is correct. 

 The wave-lengths are more accurate for no. 4 than for 3, 

 which gives a limit mean = 28996*79 + *5. The practical 

 establishment of this set excludes the connexion suggested 

 by no. 11, as 40137 cannot belong to both. 



Consider next nos. 12, 13. The difference in the means 

 shown is too great to be due to observation error. Moreover, 

 its amount, 3 '42, is very close to that produced by an oun 

 displacement in the limit, viz. 3*13. Assuming the limits 

 are then 28996*43 and 28999*56 (corrected separation) the 

 corresponding f sequences must differ by an oun multiple 

 displacement, ' The mantissa of 28996*43-18082*43 is 

 3'170020, and that of 28999*56-18240*49 is 3*192761. 



