THE MECHANICAL EQUIVALENT OF HEAT. 
421 
of our platinum thermometers (H) and the air thermometer over a range of 0° to 100°, 
and the numbers there given show that the probable limit of error at any point in the 
above range is 0°‘01 C. This limit may be further diminished by the “ plug correc¬ 
tions” (see p. 410), then neglected by us. It must be remembered that this limit 
of error refers to the actual elevation, and is probably decreased when we come to 
consider differences over a range of 11° C. Further, the table there given is a record 
of only a single series of observations, and must include experimental errors, which 
would tend to mean out if sufficient observations were taken. 
Again as regards differences, the quantity which chiefly affects us (apart from 
experimental errors, etc.) is the value of 8 in the equation 
d = e —pt = B [B/lOOf — d/100] ; 
and it is therefore important to point out to what extent the variations we have 
observed in the value of 8 may influence our results. 
We have standardized our mercury thermometer by means of the platinum 
thermometer N.* (For particulars see ‘Phil. Trans.,’ 1891, A, p. 151.) This 
thermometer contained two wires, and by placing these in series a third resistance 
could be determined ; it thus practically contained three coils, which we denote by 
A -f B, A, and B. 
The fixed points (B^, Bj, and B^) of this thermometer have been re-determined at 
regular intervals during the past two years, and no change of any consequence has 
been observed. 
The values of its constants are as followst:— 
A + B. 
A. 
B. 
Rg 
9-8636 
5-9881 
3-8762 
Ri 
13-2987 
8-0743 
5-2252 
R, 
24-2703 
14-7631 
9 - 5.343 hence 
6 
1-641 
1-645 
1-640 
Now, the value of d at 14° C. and 25° C. obtained by assuming the above values of 
8 are — ‘198 and — ‘308, and it is on the difference between these numbers ('110) 
that possible error in the value of our range depends. Assume it as j^ossible that the 
true value of 8 is as high as 1‘70 or as low as 1’60 (our greatest differences in any 
determinations of this constant have been ’009) we should then get — ‘20.5, — •319, 
* The coil of thermometer H was so long that it was barely contained in the calorimeter, and, as it 
required complete immersion, it was not suitable for use in this case. 
t The constants differ slightly from those previously published (‘ Phil. Mag.,’ December, 1892), as the 
“plug correction” (see p. 410), has, for the first time, been applied to them. By assuming that R = 0 
we get for the absolute zero — 270'66, — 270’55, and — 270'86 instead of the numbers there given. 
