776 Proceedings of the Royal Society 
it by degrees from over the hot junction we can make the cooling 
fast enough at the lower ranges. In fact, I believe that if I do 
not succeed in getting a sufficient number of practically infusible 
metals to construct my proposed thermometric arrangement, I may 
be able to make a fair approximation to temperatures by simple 
time observations made with the hot tube, surrounded by some 
very bad conductor, such as sand, where the surface in contact 
with the air is always comparatively cool, and where therefore we 
can accurately calculate the rate of cooling. 
Curves I., II., III., in the plate were drawn by means of this 
apparatus. The hot junction consisted of an iron wire, a palladium 
wire, and (for the several curves in order) — I. Hard platinum; 
II. Pt 85, Ir 15 ; III. The other alloy of Pt and Ir. The free 
ends of the palladium wire, and of the platinum or alloy, were 
joined to iron wires, and the junctions immersed in test-tubes filled 
with water resting side by side in a large vessel of cold water. 
The other ends of these three iron wires, and the wires of the 
galvanometer, were led to a sort of switch, by means of which 
either circuit could be instantly made to include the galvanometer. 
Readings were taken of each circuit as fast after one another as 
possible (with the galvanometer I employed about 6 '5 seconds was 
the necessary interval), and the mean of two successive readings of 
one circuit was taken as being at the same temperature as that of 
the intermediate reading of the other. 
The indications of these curves are very curious as regards the 
effect of even small impurities on the thermo-electric relations of 
some metals. It is probable, from analogy, that the curve for iron 
and 'pure platinum, in terms of temperature, would be (approxi- 
mately, at least ; even if it should be the iron, and not the platinum 
metal, which is represented by a broken or curved line) a parabola 
with a very distant vertex. And it appears probable that when 
the wire of curve III. is analysed it will be found to contain even 
a larger percentage of iridium (?) than that of curve II. 
I find, by tracing these curves on ground glass, allowing for the 
difference between temperatures and the indications of an Fe-Pd 
circuit, and superposing them on a nest of parabolas with a com- 
mon vertex and axis, that they can be closely represented by suc- 
cessive portions of different parabolas (with parallel axes) whose 
