45 
p. 42 with the given magnetic separation is examined, the question 
naturally rises: 
“Why do 4 3032.90 (III,) and 4 3175.12 (VIII,) occur in different 
series for Tin, though they exhibit the same splitting-up ?” 
The same question applies also for Antimony 4 2770.04 (XII,) and 
À 3267.60 (XIII,), and also for Antimony A 2598.16 (XVIII,) and 
A 2528.60 (XXII,). 
To answer this question I have traced every time two lines as 
10° À-! and examined by means of my model without giving it a 
rotation, what would be about the frequencies of the other terms of 
the series that is perfectly determined without rotation by these two 
points. In this way I have arrived at the following results: 
If we consider Sn 4 3175.12 as e=3 and Sn23032.90 as e=—4, 
we get 10°A-' — + 28400 for z—= 1, which does not agree with 
any observed line. (The nearest lines have the frequencies 27353.20 
and 30023.63). 
If we consider these lines as e=—3 and x—5, we find 
108 A! = + 32450 for «x= 4, which does not agree with any line. 
e= yields 10°4-!'= + 29500, which might then possibly be 
30023.63. But this is not very probable either, for the line which 
+ 1.79 s 
+ 1.22 p 
agrees with this (A 3330.75) exhibits a quadruplet') 0 in the 
— 1.22 p 
—1.79s 
magnetic field, and so very certainly does not belong to this eventual 
series. In this way I have ascertained that the lines in question 
cannot be ranged together with others in one and the same series. 
I have obtained corresponding results with the other lines which 
show the same splitting-up. This has rendered it very probable that 
the rule: “All the terms of one and the same series present the 
same resolution in a magnetic field’, cannot be reversed, and so 
it is my opinion that the argument that I have not ranged lines 
which present the same splitting up in the same series, cannot be 
advanced as an objection to the classification of the Tin- and Antimony- 
spectrum given by me. 
1) Purvis. Proc. Cambridge Phil. Soc. 14. 1907, p 220. 
