312 Prof. W. M. Hicks on the 



linkage effect where a small change in a line makes all the 

 adjacent links normal. I£ the corrected value is used the 

 line is 41608 77. The separation is now 68'81, which is 

 equal to 22 Sj on the limit ( = 68*88). 



In m = 3 the set are good but 22034 (in the above gap) is 

 •a line given by Duffieid and Rossi as appearing in the band 

 spectrum. Again the doublet separation appears as 75*5 or 

 as due to 24S X =7S. 



The evidence again is in favour of a small negative f. 



In the spark spectrum occurs a large block of lines between 

 17604 and 18240, all weak and of the same character with 

 intensities fluctuating between 1 and 4. There is a corre- 

 sponding block between 39622 and 40353 about the same 

 distance on the opposite side of F(cc ). The whole have the 

 appearance of the F and F type. A considerable number 

 can be arranged in F and F sets. The separations in each 

 block are very close to values produced byoun displacements 

 in the sequent. As the change per oun is so small, varying 

 here from -77 to '67, a particular agreement has no weight 

 tor that particular case. The fact that so many closely agree, 

 however, supports the general conclusion as to their belonging 

 to one system. The whole set of observed lines is given in 

 the subjoined table. In this table the lines are so arranged 

 that suitable F 2 and F 2 come on the same level in each 

 ■column. Means in the neighbourhood of 28998 and 28923 

 are entered as 98, 23 respectively, pairs giving the former 

 = F 2 x> on the same level on the left and those with the latter 

 = F 1 co entered on the right opposite the suggested Fj line 

 and the associated F line indicated by dotted lines. The 

 first thing that strikes the eye is the number of cases close 

 to 28996, only 3 near 99 and no others near them, also that 

 the cases near 21 are much more irregular, and show no 

 special concentration about any particular value. Several 

 of the instances of " 21 " are associated with lines which are 

 also associated with others for the " 96 " mean. They cannot 

 both be related as Fx and F 2 lines and there must be several 

 mere coincidences. The cause of these coincidences is clearly 

 seen when it is noticed how large a number of the lines have 

 separations of the order 72. These are indicated by numbers 

 on the left of the first column. Two such independent series 

 seem to exist. As these separations are about the same as 

 F 2 (oo ) — Fj(oo ), the frequent appearance of both " 21 " and 

 *" 96" means in connexion with the same line is to be ex- 

 pected. As, however, the " 96 " values have a close agree- 

 ment, while the observed " 21 " are irregular, it may be 



