512 



Dr. G. J. Stoney's Analysis of 



Accordingly all the observed points lie nearly on this para- 

 bola. The parabola is in the position shown in fig. 3, from 

 which it follows that the primary curve (which represents the 

 relation between n and l/?n) is a curve winch is nearly a 

 hyperbola. See PL VII. figs. 4 and 5. 



Application to Series D. 



Series D is best represented by a straight line for its derived 

 curve, and accordingly by a parabola for its primary curve. 

 Putting as before x=n and z=1000/??i 2 , and taking 



x = a — olz 



as the equation of the derived curve, in which 



a = 244'93 and loga='04357 3 



and computing the successive terms, we find: — 



m. 



n calculated. 



+ k by observation 



(which is nearly the same 



as n). 



Difference. 



1 



-860-6 



Too far in ultra-violet for observation. 



2 



- 31-17 



Too far in ultra-red for 



observation. 



3 



+ 122093 



+ 122-036 



+0-057 



4 



+ 175-834 



+ 175-884 



-0050 



5 



+200-709 



+200746 



-0-037 



6 



+214-221 



+214-256 



-0035 



7 



+222-368 



+222-363 



+0-005 



8 



+227-656 



+227-676 



-0-020 



results which show that the primary curve of Series D ap- 

 proximates very closely indeed to a parabola. See PL VII. 

 figs. 4 and 5. 



Application to Series S. 



This is the most interesting of the series so far as the present 

 inquiry is concerned, because it is the series to which belongs 

 the outlying double line which has been supposed to be a 

 satellite of one of the terms of Series D. 



