404 Proceedings of Boy at Society of Edinburgh . [july 18, 
•= a<f>(t) - aft) - + V(t).f(t) * 
= [>-F(£)] 
by the experimental result in equation (2), 
^ = [a-F(f)][a-F(«)fy(<) (0- 
Also, since F(tf), the thermoelectric power of FeZn, is known to be 
of the form p + qt, the experimental result in equation (3) may 
be written 
where n and b are independent of t, . \ we have proved (by experi- 
ment) that 
[a - F (£)] [a - F(ty\\f/(t) = n[b ± F(^)] 2 ; 
i {/(t) must be independent of t. 
And thus equation (2) becomes 
m 
Thus we have established by experiment (and the three assumptions 
noted above) that the Peltier effect is proportional to the product of 
the thermoelectric power and the absolute temperature. 
§ 5. Measurement of Change in Specific Heat. 
As the correction for increase of specific heat due to rise of tem- 
perature becomes very large at the higher temperatures, it w r as 
necessary to measure it for the metals used in the above experiments, 
A new method of doing this occurred to the writer, and, although 
he believes that much more accurate results might be got by this 
method, the results already obtained seem worth publication. The 
method consists in measuring, at different temperatures, the Joule 
effect in a strip or strips of the given metal. For this purpose two 
narrow strips (of iron, say) were doubled along the two sets of 
junctions of an FeArg thermopile (wrapped in thin asbestos paper), 
* These four terms correspond to the four sets of thermopile junctions. 
= xj/(t) - constant. 
