field are hot modified by a cavity formed in the vertical axis, as 
was usually the case, for it is clearly those portions towards the 
surface of the ellipsoid that are the chief contributors to them. On 
the other hand, they might obtain a greater ’relative influence, but 
as the observations shew, the sunt of the corrections arising from 
this cause is so small that they may be regarded as independent of 
the susceptibility within the limits of accuracy of the experiments. 
In that case this difficulty completely disappears. 
4. Dimensions of the ellipsoid. The internal volume was obtained 
by filling the ellipsoid with mercury and weighing it. It was 0.1812 c.c. 
The change of volume under atmospheric pressure was found to be of 
no account by pumping the space above the mercury free from air 
and observing the position of the mercury in the capillary. 
The external axes were measured directly. Then the thickness of 
the glass at ten different points was determined by focussing a 
microscope on the image of the outer surface formed on the mercury 
with which the ellipsoid was filled. It changed but slightly from 
place to place. The mean was taken and twice that value was sub¬ 
tracted from the external measurements. The results were : 
1.044 cm. 
and 
0.335 cm. 
Calculating the volume from these figures we get 0.1925 c.c. which 
is about 6 °/ # greater than the true volume as directly determined. 
This is accounted for by the special shape of the meridian section 
which curves somewhat too strongly at the outer ends. For calcu¬ 
lating the coefficients of demagnetization we took a mean ellipsoid 
with the same major axis and the minor akis small enough to give 
the real volume 1 ). The data for the calculation were therefore: 
1.044 cm. 
and 
0.3173 cm. 
5. Opposing couple. The suspension spring and the stretching 
wire were the same as were used for the liquid oxygen. We must, 
however, allow for the rubber supply tube for the oxygen. This 
(which was chosen as thin as possible) modified both the zero and 
the constants of the total opposing couple, as soon as the pressure 
i) It is clearly not quite right to do this; there are, however,experimental data 
to support this method of correcting: V. Quittner (Diss. Zurich 1908, also Arch, 
sc. phys. et nat Geneve, Sept.—Nov. 1908) found that this method of treatment 
was sufficiently accurate even for discs, bodies that deviate far more from an 
ellipsoid than those we used. 
