1887.] Mr A. Campbell on Measurement of Peltier Effect. 383 
tightly in asbestos wool in the middle of a series of copper boxes 
or tubes, separated from one another by asbestos packing. A 
thermometer was inserted with its bulb as near to the junctions as 
possible. By properly arranging Bunsen burners underneath, it 
was possible to maintain the junctions at a constant and uniform 
temperature. Throughout the whole series of experiments, the 
mercury thermometers employed were regularly compared (at the 
ordinary temperature) with a standard one which was never heated 
up, and which had been compared with a Kew standard. All the 
temperature readings were thus corrected. 
The following was the usual mode of observation: — When the 
temperature had become sufficiently uniform (which was often not 
until several hours after the heating was begun), the current was 
sent in the positive direction for 30 seconds, then broken for 30 
seconds, put on in the negative direction for 30 more, broken for 
30 more, and so on, for about 10 minutes. The deflections of the 
mirror galvanometer (connected with the measuring thermopile) 
and of the Helmholtz galvanometer were observed simultaneously 
at the end of each period of 30 seconds. If the scale-readings of 
the mirror galvanometer be a ly a 2) a s , a 4 , a 5 , (a x being 
the original zero-point), then the average of — a 2y a 4 — a 3 , - a 6 — « 5 , 
....... is taken as the measure of the temperature difference 
caused by the given current (in 30 seconds). 
Since the specific heats of most metals increase considerably as 
the temperature rises, it is necessary, in comparing the Peltier 
effects at two temperatures, to make a correction for the change in 
specific heats. In the absence of definite measurements, this in- 
crease in specific heat was taken as approximately =y^o per degree 
centigrade for iron and nickel, and =yoVo P er degree for cadmium, 
zinc, and German silver. This correction was reckoned from 20° C. 
The correctness of the method was tested by applying it to prove 
that the Peltier effect is proportional to the strength of the current. 
As was to be expected, the temperature nearly always rose slowly 
during the ten minutes’ observations (owing to the Joule effect); 
a small correction was applied for this. Table I. gives some of the 
measurements of the Peltier effect for different current-strengths. 
The first three sets are for iron-cadmium, and the fourth for iron- 
zmc. 
