458 Prof. W. M. Hicks on 



the measures of Meggers* and Meissnerf give 248*43, Eder 

 and Valenta 248*42, whilst Meggers' estimates for his 

 probable errors for the two lines give for dn *03 and *01, 

 probably less on a difference. The separation for D(2) as 

 calculated below from Fabry and Perot's interferometer 

 measures of D n , D 22 is 248*44. The separations for the S 

 and D doublets therefore agree within one unit in the second 

 decimal. Now a B displacement on the limit alters the 

 separation by *06 and the limit itself by 4*94. Consequently 

 the two limits S (co ), D (cc ) cannot possibly differ by more 

 than a shift due to one oun, i. e. by 1*2, and are practically, 

 if not absolutely the same. If, however, the observed sets 

 for m = 4 as shown in Tables I., II. be taken as normal, 

 the limits as calculated from the first three lines are 

 S 1 (oc) = 31536*29±2 and D 3 (oo ) = 31515*48±10, and they 

 cannot be the same. At least one set of the fourth order 

 must be abnormal. In the case of Si (4), Kayser and Runge 

 and Eder and Valenta (spark) agree and Crew and Tatnall 

 give d\ = 'Q2, whilst in S 2 (4), C. T. give d\ = -0± on K. R, 

 and E.V. have not observed it. We may be justified 

 therefore in taking observation errors as small. The ob- 

 served separation (K. R.) of 248*71 is thus abnormal. It is 

 •27 too large by K. R. and *13 by C. T. J The observations 

 are sufficiently exact to prove that a negative displacement 

 has taken place, but not so exact as to determine how much. 

 If the displacement be — xh, x—1 or possibly 2, v is increased 

 by '06 x and the limit by 4*94#. In other words the ob- 

 served lines must be regarded as being affected by a limit 

 4"94# larger than the others. Treated in this way the true 

 limit Si (cc ) calculated from m = 2, 3, 4 becomes 31536*29 — 

 12*46 .^±2. The data for the D give a wide margin of 

 variability for D(go ), chieflv due to uncertainty in D x (4). 

 K. R.'s measures give n = 27103*00, C. T.'s 27104*62 with 

 separation 252*14 instead of a value less than 24844. Also 

 E. V. give a spark line n = 27 109*23 which gives a separation 

 of 245*66 with D 22 (4), or a satellite separation of 2*78. The 

 data are too uncertain to secure a closer value to the true 

 value of D(oo). It remains to see what evidence can be 

 obtained from summation lines. 



The material from the summation lines is given in Tables 

 I., II. The S give direct indications from S x (2), S 2 (3), S 2 (4) for 

 a limit 31523*47 + 2*5; .. 23*48 ±*7; .. 24*82 + 1*6 and from 

 the S x (4) of ..28*35+ 1*5. But we have already seen that 



* Bureau of Standards, Washington, No. 312 (1918). 

 f Ann. d. Phys. 50. p. 713 (1916). 

 t But Hasbach gives v= 248-42. 



