in Alloys of Bismuth and Tin. 567 
Although the greatest care was taken in preparing the rods, 
it was found in practice that the temperature of a given point 
of a rod when this was traversed by a current varied some- 
what with the direction of the current when no temperature 
gradient was applied to the rods. 
The magnitude of this effect varied from point to point, 
and may be attributed to slight irregularities in the compo- 
sition and diameter of the rods. It was eliminated by 
measuring the change of temperature produced by a reversal 
of the current, both when a temperature-gradient was applied 
to the rods, and when these were at the same temperature 
throughout, and applying the latter, assumed proportional to 
the absolute temperature, as a correction to the former. 
Another series of observations has then been taken with 
the rods reversed, so that the temperature-gradient was in 
the opposite direction along these, and the correction applied 
as before. As will be seen from the tables given, this cor- 
rection varied from zero to about 5 per cent. of the whole. 
To obtain the temperature-gradients along the rods, the 
temperatures were measured at various points. For this 
purpose a number (usually 5) of thermocouples of copper and 
german-silver were attached at intervals along each rod, 
one junction of each being fastened to the rod, from 
which it was separated by thin strips of mica, the other being 
free outside the covering of cotton-wool, so that it could be 
immersed in a vessel of water. 
Each couple was then joined up in turn with the galvano- 
meter, and the water, in which the second junction was 
immersed, heated until there was no current in the circuit 
when this was completed. 
The temperature of the water, as shown by a thermometer 
reading to tenths of a degree, was, then, the temperature of 
the rod at the point to which the junction was attached ; and 
by plotting temperatures against lengths along the bar, the 
value of the temperatur e-oradient at any point could be found. 
Readings of the temperatures were taken both when there 
was no current passing through the rods, and when these were 
heated by the various currents used in the experiments. 
The only other datum required is the resistance per unit 
length of the rods ; when this has been found by measuring 
the resistance of each rod and taking the mean of the two 
values, the specific heat of electricity is calculated from the 
expression given above (p. 566). 
