OF CRYSTALS AND OTHFR BAD CONDUCTORS. 
483 
The cans at the extremities of the bar were supported on slides attached to the 
framework in such a way that the bar could be moved parallel to itself by screws 
passing through the framework and bearing in conical indentations in the backs of the 
cans in the prolongation of the axis of the bar. These supports were sufficient while 
the bar was entire, and the experiments to determine its conductivity were being 
made. After the bar wms cut for the insertion of the discs, it became necessary to 
support the cut ends in such a way as (1) to enable the motion of the bar parallel to 
its axis still to take place ; (2) to enable the cut faces to be adjusted parallel to each 
other ; (3) to cause as small as possible a disturbance of the isothermal surfaces. 
These objects were secured by supporting each free end in a loop of thin string, 
hanging from the upper part of the framework. The ends of this loop were attached 
to screws for the purpose of raising or lowering it as required. The sides of the loop 
made an angle of 60° or 70° with each other at the bar. These loops were sufficient 
to support the bars properly if the arrangement was kept horizontal. It was, 
however, found that the mercury contacts could be made with much greater certainty 
if the bars were vertical and the mercury surfaces horizontal. On this account two 
other loops were attached to the lower part of the framework and passed over the bars. 
These were elastic, and served merely to keep the upper loops taut, and the bars in the 
same positions with respect to each other, whether the framework was horizontal, as 
in the experiments, or vertical, as in making the mercury contacts. 
Measurement of Temperature. 
The most direct method of measuring the temperature along the bar is to sink 
thermometers into holes in the bar, but this method, although it has been used by 
Forbes, Tait, and Mitchell for their large bars, is very objectionable, as it 
diminishes the available area of flow, and therefore makes the change of temperature 
along the bar more rapid than it would be for a continuous bar. The method first 
3 Q 2 
