( &i4 > 
(0,05%) in the measurements with the cathetometer of the displace¬ 
ments of the level, and from the accuracy of the measurements of 
the field-strengths, of which a discussion is given above. It is certain 
that the cause of these deviations must arise from a source other 
than the measurement ox these two data, though we cannot with 
certainty indicate what this may be. 
We may in the meantime remark that, at least in the case of the 
first series, the unsteadiness of the apparatus in the vertical direction 
in the not quite homogeneous field, and the slight inconstancy of 
the temperature have certainly been contributory causes of these 
deviations, since the second apparatus which was improved exclusively 
in these directions gave much more regular results. This remark, 
however, does not seem to account sufficiently for certain appreciable 
changes that occurred without any noticeable corresponding irregularity 
in the pressure or in the convection current of the liquid, while 
there was also no noticeable change in the shape of the meniscus. 
Liquid oxygen II. 
§ 3. Measurements by the method of the maximum couple exerted 
upon an ellipsoid. Further comparative measurements for liquid 
oxygen at various temperatures were obtained by means of the 
method of the maximum couple exerted by a uniform field upon an 
ellipsoid. This method has already been described and discussed in 
connection with the research on ferro-magnetic substances 1 ); it will 
be sufficient to discuss the modifications which were found to be 
necessary owing to the particular circumstances under which the 
method had to be applied to the present research. 
In the first place on account of the small value of the susceptibility 
it was necessary to make the couple to be measured as large as 
possible; with this end in view we chose an oblate ellipsoid of 
revolution, instead of a prolate; its axis of revolution was placed 
horizontal in a field which could turn round a vertical axis. 
The ratio that is taken between the axes is not a matter of in¬ 
difference; for a given major axis the couple, which is proportional 
t 0 (7%— NJp, is a maximum for a ratio of the major to the minor 
axis that is only slightly smaller than 3; we have therefore taken 
this value of the ratio for the construction of the ellipsoids. 
We used the same electromagnet as served for the measurements 
made by Weiss and Kamerlingh Onnes (loc. cit.). Two pairs of pole- 
i) P. Weiss, J. de phys. (4) 6 (1907) p. 655. P. Weiss and H. Kamerungh Onnes, 
Comm. N'. 114 These Proc. Jan./Febr 1910. 
