290 
necessitated a more consistent calibration. After repeated efforts to obtain 
a constant deflection for a given difference in temperature the method was 
abandoned as not being sufficiently accurate. 
The following scheme was then adopted: 
A battery of known KE. M. F. was connected in series with two resist- 
ances, R., which was approximately 100,000 ohms, and in later experiments 
kept constant, and Ri was varied from one to eight ohms to suit the condi- 
tions of the particular observation. B is a carboy in the center of which 
one junction of the couple was located. A is a second junction which was 
kept at a constant temperature. G is a galvanometer in series with the 
couple and Ri. The air in B was compressed as before and allowed to cool 
to the temperature of the bath. IK, was then closed, then the valve was 
opened to the atmosphere and immediately K, was closed and the direction 
of the deflection of the galvanometer noted. ‘This process was repeated, 
varying R: until a resistance was found such that on closing IK, there was 
no deflection of the galvanometer, until the air began to warm after the 
adiabatic expansion. This balanced condition meant that the P. D. across 
It; just balanced that due to the difference in temperature of the two junc- 
tions of the thermo-couple. 
In practice it was found better to set Ri at a given place, e. g., 5 ohms, 
and then vary the original pressure until a balance was obtained. In some 
of the earlier observations R. was varied to secure a balance, but since it 
was not known to a sufficient degree of accuracy, the other method was 
used. It then remained to calibrate the thermo-couple. This was done by 
placing one junction, encased in a jacket, in a constant temperature bath, 
and the other, similarly encased, in a bath whose temperature was varied 
till a balance against a given resistance, Ri, was obtained. The difference 
in temperature of the two junctions was then noted. Ri was again varied 
