of Edinburgh^ Session 1871-72. 
777 
tangents coincide at the points of junction, though the curvature 
is necessarily not continuous from one to the other. Hence, as at 
least a fair approximation to the electro-motive force in. terms of 
difference of temperature in the junctions, we may assume a para- 
bolic function, which up to a certain temperature belongs to one 
parabola, then changes to another without discontinuity of direc- 
tion, and so on. 
Hence either the iron, or the hard platinum and the platinum - 
iridium alloys, will be (approximately, at least) represented on my 
form of Thomson’s thermo-electric diagram ( ante p. 601) by broken 
lines, of which the successive parts are straight. This, contrasted 
with the (at least nearly) straight lines for pure metals, seems 
to show that some bodies take successively different states ( i.e ., 
become different substances ) at certain u critical ” temperatures, re- 
taining their thermo-electric properties nearly unchanged from one 
of those critical points to another. 
The curve marked IV. in the figure was obtained by plotting 
against each other the simultaneous indications of the alloy of curve 
III. and iron, and of the alloy of curve II. and iron, so as to avoid 
any disturbance from possible peculiarities of palladium. Then, to 
obtain an idea of the share taken by iron in the results, it was found 
that the electro-motive force in a circuit formed by the two alloys, 
or by either with hard Pt, is (for a very great range of temperature) 
sensibly proportional to the temperature difference of the junctions. 
The same result is easily seen from the plate, if we notice that 
the difference of corresponding ordinates in any two of curves I., 
II., III., is nearly proportional to the corresponding abscissa. Now, 
it seems a less harsh supposition that the lines representing pla- 
tinum and its alloys are nearly straight and parallel, while that of 
iron is a broken line, than that the latter should be straight and 
the former all broken at the same temperatures. On the other 
hand, this latter hypothesis would make k alternately negative and 
positive in iron, while the former would only require the platinum 
metals to have values of k alternately less and more negative than 
that of iron. 
1 may add that none of the above-mentioned effects can be due 
to altered electric resistance of the heated junctions, because the 
galvanometer resistance was about 23 B. A. units, while that of the 
