400 Proceedings of Royal Society of Edinburgh. [july 18 , 
as the temperature rose. As it was the E.M.F.s that were to be 
compared, the indications of the FeZn piles had to be corrected for 
this difference of resistance (and also for the change in this differ- 
ence). In (a), as the piles had both such small resistance, this 
correction was not required. 
The specific heat correction was deduced from the measurements 
described below. Unfortunately no measurement was made in the 
case of zinc ; so the change in specific heat for it was assumed as 
TO % per degree C. If , s 2 , s 3 be the specific heats of Fe, Arg, and 
Zn ; u\, iv 2 , w 3 the relative weights of the thermopile strips of these 
metals; a 1 , a 2i a 3 the respective percentage increments of the 
specific heats (per degree C.) ; then the complete correction (% per 
deg. C.) has been taken as 
2a 1 w 1 s 1 + a 2 w 2 s 2 -f a 3 w 3 s 3 
2w 1 s 1 + w 2 s 2 + w 3 s 3 . 
This gives 
T04 % per degree C. at 110° C., 
T08 % . „ at 120° C., 
and so on up to 
T36 % per degree C. at 180° C. 
From the values within this range of temperature those at lower 
temperatures were found by extrapolation. The effect of the small 
amount of brass and solder in the junctions was neglected. The 
specific heat of German silver was taken as TO. 
In Table X. are the results obtained with arrangement (a), and in 
Table XI. those with arrangement ( b ). In the measurements 
marked X the current was sent through the FeZn, and the Peltier 
effect measured by the FeArg ; in those marked Y the current was 
sent through the FeArg and the Peltier effect measured by the 
FeZn. The third column gives the observed values of ~ (corrected 
Ct 
as above), where D is the mean thermopile deflection (mean of 6), 
C the mean current, and t the absolute temperature. 
From the following tables it is clear that (within the limits of 
experimental error) — has the same value in the X as in the Y 
D 
measurements ; but the values of — — found do not lie on the curve 
\jt 
