436 Peirce and Willson—Temperature Variation of the 
tion of the point at which the temperature is taken, and there 
can be no question, we think, that this compels the use of ther- 
mal junctions for that purpose. We did not attempt to employ - 
fine platinum resistance coils, on account of difficulties which 
arise from the use of leads of considerable dimensions. 
In the case of certain substances, such as sulphur or cement, 
which can be made to solidify about a system of stretched wires 
carefully spaced, it is possible to imbed thermopiles in fairly 
homogeneous prisms with the leads lying in what will be iso- 
thermal surfaces, so that there need be no fear of direct con- 
duction of heat to or from the junctions, but such a process as 
this is manifestly inapplicable in the case of stone. After dis- 
cussing the possibility of boring holes from the edges of a stone 
slab, we determined that it would be worth while to try the 
experiment of building up the prism out of separate slabs, 
clamped firmly together by means of the outside plates, with 
thermopiles fitting closely in fine grooves in the faces of the 
slabs, made with the help of a milling machine. It would not 
be easy, however, to make such grooves in glass plates, and we 
saw no better way of proceeding than to clamp the thermopiles 
between the plates. This, of course, raises some serious ques- 
tions as to the difference between the indications of the the: mal 
junctions and the true temperatures at the two surfaces with 
which the wire is in contact. It is obvious that the spaces 
between the slabs should be as small as possible and therefore 
that the thermopiles should be of fine wire. In advance of any 
experiment on the subject, it seemed that the communication 
of heat from radiation alone across a film of air, at most a 
small fraction of a millimeter im thickness, would be sufficient 
to prevent any great inequality of temperature, and that the 
thermopiles would give a close approximation to the mean of 
the temperatures at the two surfaces. 
On these considerations we decided to make preliminary ex- 
periments in this manner, using marble as the experimental 
substance. How far the event has justified these considerations 
will appear later. The apparatus used in our first experiments 
was as follows: 
Upon a heavy table was placed the hot chamber A [figure 1] 
connected with the copper boilers B, B’, for generating steam. 
The boiler B, which held about 40 liters, was constantly heated 
and was refilled when nearly empty without stopping the con- 
stant flow of steam, by filling B’, heating to the boiling point, 
pee closing the opening at e with a weight which acted as 
a safety valve, and opening the communication at f. The gen- 
eration of steam in B’ served to drive the water over intu B 
without causing a cessation of the flow of steam through the 
hot chamber, and the communication between the two boilers 
was then cut off. This process was necessary at least once in 
