128 
Proceedings of the Eoyal Society 
From the first two of these means we find that the specific heat 
at 15° is 0 f 109 nearly, and that it increases by T -|^th for each 
degree. 
Now, Forbes’ experiments on iron indicated that the quantity 
Jc 
the ratio of the conductivity to the specific heat, diminishes by 
0 
about —-^th part for each degree from 0° C. to 200° C. Hence 
it is clear that, in this case at least, the alteration of specific heat 
cannot be neglected in estimating that of conductivity. For it 
follows from the numbers just given that the diminution per 1° in 
the conductivity of iron is really only about —^th of the whole 
amount. My own experiments with Forbes’ bars gave an average 
Jc 
change of - less than that due to the increase of c alone, thus indi- 
cating an increase of conductivity with rise of temperature. Ang- 
strom’s result, on the other hand, is considerably greater than that 
of Forbes. But the range of temperatures he employed was not 
above forty degrees. For reasons pointed out in my paper above 
Jc 
referred to, I consider Forbes’ estimate of the value of from 0° to 
’ c’ 
150° C., to be probably very near the truth. In other metals the 
change of specific heat is usually less than in iron. But so is also 
Jc 
that of - . It would thus appear that we cannot yet state positively 
c 
that there is any metal whose conductivity becomes less as its tem- 
perature rises ; and thus the long-sought analogy between thermal 
and electric conductivity is not likely to be realised. 
In the method devised and carried out by Forbes, the change of 
specific heat must be attended to during the calculations. Thus we 
cannot, without going over again the whole numerical work con- 
nected with what he called the Statical Curve of Cooling , estimate 
accurately what will be the effect of this element upon the values 
of the conductivity. But we can easily show that its influence 
upon Angstrom’s results is to be calculated, at least approximately, 
by the simple process above. 
To avoid the error introduced by supposing rate of surface loss to 
be proportional to v, we take (instead of a bar) a plane slab heated 
and cooled periodically over one surface. 
