4000 
3000 
96 
hence a similar result as for the first method. (See diagram II), 
The deviations inter se are now much larger, as was, indeed, to be 
Di agram II. 
0.4 0.5 0.6 0.7 0.8 0.9 1.0 ge 1.2 1.3 14 1.5 
expected, as on the small intervals 4,—A, the inevitable errors in 
/, (an experimental quantity!) make themselves very greatly felt. 
Thus Sie give an imaginary value for 7, but when for 
| eed 
j= 92,0 . J, == 22 is ‘taken instead ‘of Zp 19, then. Ze would 
become == 18000°. 
In this manner particularly the smaller values of J, are unfa- 
vourable, hence the values for 2, == 1,5 and 2, =1.8 are not much 
to be trusted. 
The values of 4, for 2< 0,5 are strictly speaking also unreliable, 
because the graphical interpolation — as indeed every other too — 
becomes very inaccurate here. : 
When we leave all these doubtful values of 7’ out of conside- 
ration we come to the result that particularly in the region of the 
reliable values of 7’ (the full line in the diagram) there is an un- 
mistakable tendency of T to decrease on the increase of 4, hence 
exactly the reverse of what we thought we might expect a priori. 
In a following paper I propose to discuss the question to what 
this unexpected result is to be attributed. 
Utrecht, March 1919. 
