Value of the Silver Oun. 315 



Their difference is 22741, whilst 54S is 22737, so that the 

 condition is fulfilled. If then 18082 is denoted by F 2 (.r) the 

 set are F 2 (tf), F,(a>), (-8,) F^.r) (545), (-8,) F 2 (#) (543). 

 The two sets give for F 2 (oo ), 28996'43(*5) and 28996'72(-9). 



Nos. 12 and 5 are mutually exclusive. Since the above 

 evidence for the real connexion of 12, 13 is so good, we 

 should perhaps omit 5, which indeed would give practically 

 the same limit, viz. 96*24(*5), although if* it is a mere ques- 

 tion of probability both might be taken. 



All the separations in (1, 2) deviate considerably from any 

 normal values. If (1) is a real set the F(co) and F(go) 

 must be different as well as the sequents, and it is of no use 

 for our present purpose. No. (2) gives the limit mean so 

 close to the others that the set is almost certainly correctly 

 allocated, and it may be used for the determination of f. 



In (7) the limit mean would seem to be analogous with 

 that of (13) and will be treated on that basis. Using then 

 (2,4, 7, 8, 12) as of equal weight and (3, 9, 10, 13) of half 

 the weight of the former, the most probable value of the F 

 limit is 2899643, and if (5) be included is 28996*41. If tbe 

 possible errors in the individual measures are not greater 

 than those indicated, the possible variations of the true limit 

 from 28996-43 are +*24 and -*14, that is <*3. 



The value calculated from 34A for F„co was 



28998-98 + l-286f--10? 1 -p 2 ). 

 Hence l'286f = -2"55 + -3+ 'l{pi—p a ), 

 f=-l-98±-2 + -08(^-p 2 ) 



= -l-98±-3. 



The value of f as determined above from the D and D 

 material was £= — 1*90 + *4. Both agree within limits. We 

 may therefore take it as established that £= — 2 + *3, It may 

 be noted that the best reading in the D, that from Di(2), 

 Djl (2) gives f = —2*02. This definitive value of f determines 

 A with great accuracy. Now 



A = 27786-57-l'267£ + -lQt> 1 - / > 2 ) 

 = 27789-10±-3. 



This has been determined on the basis of the accurate value 

 of the doublet separation combined with a very close approxi- 

 mation to f from the preceding considerations. There is, 

 however, another quite independent method of arriving 

 at the value of A, viz. by deduction from the fact that 



